Bigger, Faster, Stronger: Implications of Inter-Species Interactions for IRM of Lepidopteran Pests of Bt Maize in Africa

A hypothetical scenario of mixed populations of Busseola fusca (Fuller) (Lepidoptera: Noctuidae), Chilo partellus (Swinhoe) (Lepidoptera: Crambidae), and Spodoptera frugiperda (J.E. Smith) (Lepidoptera: Noctuidae) was used as a model to investigate the potential effects of mixed populations of lepidopteran pests, on the design and implementation of insect resistance management (IRM) strategies for Bt maize (L.) (Poaceae) on smallholder farms in Africa. To predict the structure of such mixed populations in different agroecological zones, the biological and behavioral characteristics that affect the competitiveness of these species were identified and analyzed. Additionally, the validity of the assumptions that underlie the high-dose/refuge strategy was compared among the three species. Differences between the species, and the influence thereof on the choice of IRM strategy for a specific environment, were explored through analysis of three hypothetical scenarios. We suggest that the use of separate refuges as a component of an IRM strategy against mixed pest populations in smallholder Bt maize fields may be unwise. A seed mixture approach, coupled with an effective integrated pest management (IPM) strategy, would be more practical and sensible since it could limit the opportunity for a single species to dominate the species complex. The dynamic interactions in a multi-species community and domination of the species complex by a single species may influence moth and larval response to maize plants, which could lead to an increased infestation of Bt plants, and subsequent increased selection pressure for resistance evolution. This article provides insights into the unique challenges that face the deployment of Bt maize in Africa.


Fig. 1. (A) Chilo partellus and (B)
Busseola fusca larvae (4th and 5th instar, respectively). The appearance of leaf feeding (C) and stalk boring (D) damage for both species is similar (Photo credits: A. Visser; H. du Plessis). beneficial insects (e.g., parasitic wasps that attack stem borers) become more abundant, which further contributes to pest control and the aims of IPM (Wolfenbarger et al. 2008, Romeis et al. 2019. Reduced larval feeding damage to ears of Bt maize plants also limits the formation of mycotoxic fungal infections (Munkvold et al. 1999;Hellmich et al. 2008;Ncube et al. 2017Ncube et al. , 2018Pellegrino et al. 2018).
Unfortunately, this technology's greatest strength-the seasonlong expression of highly effective Bt toxins-is also the source of its greatest weakness, i.e., evolution of resistance in pest populations, due to the sustained selection pressure exerted by Bt toxins (Tabashnik 1994, Glaser and Matten 2003, Siegfried and Hellmich 2012, Siegfried and Jurat-Fuentes 2016. The threat that insect resistance evolution holds to the sustainability of Bt crops prompted the development of insect resistance management (IRM) strategies before the commercial release of the first Bt crop (Gould 1998, Tabashnik andCarrière 2017). The goal of these strategies is to delay resistance evolution in target pest populations (Head andGreenplate 2012, Onstad et al. 2018). Several IRM strategies have been proposed, but the most popular has been the high-dose/refuge (HDR) strategy (Gould 1998, Bourguet et al. 2005, Tabashnik et al. 2013). This approach involves planting a refuge area of non-Bt maize plants to accompany Bt maize plantings, with the aim of sustaining a susceptible population of the target pest. The large numbers of homozygous susceptible individuals are then the primary mates for the rare homozygous resistant individuals that survive on Bt plants, giving rise to predominantly heterozygous offspring (USEPA 1998, Gould 2000. The expression of the Bt toxin should also be of a high enough dose (referred to as high-dose expression) to control the heterozygous individuals, leading to an overall reduction in the frequency of resistant alleles in the pest population (Gould 2000;Bates et al. 2005;Tabashnik et al. 2008Tabashnik et al. , 2009Tabashnik and Carrière 2017).
The HDR strategy is based on several key assumptions about the target pest population and the GM crop, i.e., 1) initial low frequency of resistance alleles in the pest population, 2) fitness costs are associated with resistance, 3) resistance is incomplete, 4) resistance must be functionally recessive, and 5) presence of abundant refuges of non-Bt host plants (Tabashnik and Carrière 2017, Onstad et al. 2018, Anderson et al. 2019. Violation of these assumptions leads to rapid evolution of resistance (Tabashnik et al. 2013), e.g., B. fusca populations in South Africa (Campagne et al. 2013(Campagne et al. , 2016Van den Berg et al. 2013). Several cases of field-evolved resistance against Bt crops have been reported worldwide (Tabashnik and Carrière 2017), mostly due to poor compliance to resistance management strategies (Kruger et al. 2009(Kruger et al. , 2012aTabashnik et al. 2013;Farias et al. 2014a,b;Tabashnik and Carrière 2019).
A mere 8 yr after the commercial release of the single Bt toxin (Cry1Ab) maize event MON810 in South Africa, the first case of B. fusca resistance was reported (Van Rensburg 2007, Kruger et al. 2011. Resistance to Cry1Ab has since been documented in many maize production areas in South Africa. Currently, the growers in South Africa are starting to use a multi-toxin (pyramided) maize event (MON89034), which simultaneously expresses the Bt proteins Cry1.105 and Cry2Ab2 (Strydom et al. 2018). This is the only pyramid event released commercially in South Africa and it is still effective at controlling Cry1Ab-resistant B. fusca populations (Strydom et al. 2018). The pyramid event started to replace the single-gene event during 2012, but cultivation of both events still takes place in most of the maize production regions of the country. Strydom et al. (2018) did report, however, that a shift in susceptibility of B. fusca to the pyramid event was observed.
The rapid evolution of resistance of B. fusca populations in South Africa highlighted the importance of an effective IRM strategy to ensure the sustainable use of Bt maize in other African countries . However, the design of an IRM strategy for the Bt maize receiving environments on the continent is complicated by several factors. Agriculture in Africa is practiced both on large-scale commercial farms and smallholder farming plots, with the latter being by far the most prevalent (Thompson 2008, World Bank 2008, Aheto et al. 2013. This heterogeneous nature of the agricultural system makes the implementation of a standardized, structured IRM strategy (as is the case in developed countries) unrealistic (MacIntosh 2009, Jacobson andMyhr 2012). Implementation of separate refuges on smallholder plots is challenging, due to the limited scale of production (Bates et al. 2005, Carroll et al. 2012, Assefa and Van den Berg 2015, Fisher et al. 2015, Kotey et al. 2016, Li et al. 2017. The suggestion has been made that the use of seed mixtures, also known as refuge-in-a-bag or seed blends, could be a workable alternative to planting separate refuges in smallholder agricultural systems (Carroll et al. 2012, IRAC 2013, Carrière et al. 2016). This approach involves mixing, at a predetermined ratio, the non-Bt seed with Bt seed. This means that the refuge plants are incorporated randomly within a field of Bt maize, simplifying compliance to refuge requirements (Gould 1998, Carrière et al. 2016. Apart from being easier to implement, using seed mixtures and pyramids for IRM have the advantage of being more resilient to certain violated IRM assumptions (Storer et al. 2012). For example, a pyramiding strategy could still delay the evolution of resistance even if only one type of heterozygote has high mortality (Roush 1998), and seed mixtures could still be successful despite the relatively high frequency of resistance alleles and a lack of recessive inheritance of resistance (Caprio 1998, Zhao et al. 2003). Unfortunately, the success of the seed mixture strategy is contingent on assumptions of its own, which chiefly relates to the movement behavior of target pest larvae from plant to plant (Mallet and Porter 1992, Burkness et al. 2015, Onstad et al. 2018. For ear-feeding lepidopteran species, there is the additional concern that Bt pollen contamination of the non-Bt ears creates a mosaic of kernel tissue on the Bt ears where a mixture of low-dose Bt kernels and kernels with two or more Bt proteins occur adjacent to non-Bt kernels (Chilcutt and Tabashnik 2004, Burkness et al. 2011, Burkness and Hutchison 2012).
An effective IRM strategy must, therefore, take into consideration both the practical limitations of the agricultural system that it intends to serve, but also the behavior and biology of the target pest (Gould 1998, Head andGreenplate 2012). This is complicated; however, when the target pest is part of a mixed population of different pest species (Roush 1997, Bates et al. 2005). An example of this is where B. fusca is found not only in a maize stem borer complex but also with other leaf-feeding lepidopterans (Ogol et al. 1999;Van Wyk et al. 2008;Mutamiswa et al. 2017b;Ntiri et al. 2016Ntiri et al. , 2019Van den Berg et al. 1991a, b;Krüger et al. 2008;Ong'amo et al. 2013;Kuate et al. 2019). Van Wyk et al. (2008) reported 12 lepidopteran species to occur on maize in South Africa, although only a few have pest status. Pest species generally differ significantly with regard to biological and behavioral aspects, even if they are closely related, such as the stem borers and leaf-feeding Lepidoptera. Such differences could complicate the design of Bt maize IRM strategies in areas where mixed pest communities occur (Bates et al. 2005).
Our goal is to review the potential effects of mixed populations of lepidopteran pests on the design and implementation of IRM strategies for Bt maize on smallholder farms in Africa. This thought experiment focuses on mixed populations of the two most important stem borer species, B. fusca and C. partellus, and the invasive leaf feeder, S. frugiperda. In the first part of this article, we compare the biological and behavioral characteristics that affect the competitiveness of these species in order to predict which of them may dominate (Fig. 3A) in different agroecological zones (Fig. 4). In the second part, we review the assumptions that underlie the HDR strategy and compare the validity of these assumptions for each of the three above-mentioned species (Fig. 3B). Finally, the effects that differences in species biology and composition of pest population could have on the suitability of an IRM strategy for a specific Fig. 3. Summary of factors that impact the choice of IRM strategy to implement along with the cultivation of a Bt crop. (A) Receiving environment, which refers to the agricultural practices of the individual farmers, the level of structure of the wider of the agricultural system, and the agroecological zone where the Bt crop is to be cultivated. (B) Validity of the IRM assumptions regarding the genetic and behavioral characteristics for each pest species within the pest complex that will be a target of the Bt crop. (C) The competitive interactions that form the basis of the structure of the pest complex. The competitive interactions between pest species will have a significant impact on the behavior of individual pest species.
receiving environment is explored through discussion of three hypothetical scenarios (Fig. 3C).

Competition Among B. fusca, C. partellus, and S. frugiperda
In a mixed population of insect pest species, it is the direct and indirect competitive interactions that determine the structure of the population (Reitz and Trumble 2002, Wootton and Emmerson 2005, Kaplan and Denno 2007. Two categories can be used to distinguish between competitive interactions: interference and exploitative competition (Fig. 5). Interference competition can manifest in direct and indirect interactions, with direct interference involving the infliction of physical harm by one individual to another (through cannibalism, fighting, and killing).
Indirect interference, on the other hand, refers to the use of repellent chemicals, marking of plants, or aggressive behavior to ward off other individuals from a resource (Fig. 5A) (Denno et al. 1995;Duyck et al. 2004;Mutyambai et al. 2015Mutyambai et al. , 2016Ntiri et al. 2017). Conversely, exploitative competition does not include any direct physical interactions but rather denotes the deprivation of access to specific resources of one species by exploitation of the resources by another (Fig. 5B) (Denno et al. 1995, Reitz and Trumble 2002, Preisser and Elkinton 2008. Therefore, in a mixed population, the dominance of either B. fusca, C. partellus, or S. frugiperda will be determined by how effectively they compete directly and indirectly for resources such as food and space. The first report of the presence of C. partellus in Africa was made in Malawi during the early 1900s (Tams 1932). This species had since expanded its range throughout east and southern Africa, initially keeping to warmer lowland areas before it also invaded higher altitude regions (Bate et al. 1991, Kfir 1997, Ebenebe et al. 1999, Yonow et al. 2017. Its expanding distribution has brought it into competition with indigenous stem borer species, with C. partellus often displacing the native species to become the predominant pest of maize and sorghum (Sorghum bicolor) (Kfir 1997, Van Rensburg and Van den Berg 1992, Van den Berg and Van Hamburg 2015, Mutamiswa et al. 2017a). However, the recent arrival of another invasive species, S. frugiperda, could influence inter-species interactions within stem borer communities, a possibility highlighted by Ntiri et al. (2019). Denno et al. (1995) found that competition was higher when phytophagous species share a feeding niche, especially internal/concealed feeding niches (e.g., stem borers) where they utilize the same space-and time-limited resources. The occupation of the whorl as a niche of S. frugiperda larvae could mean that there will be minimal direct competitive interference interactions with stem borer larvae, except for the initial developmental stages during which stem borer larvae also feed inside the whorls of maize plants (Berger 1992, Slabbert and Van den Berg 2009, Calatayud et al. 2014. However, because the mixed-species complex shares the resources (maize plants) within a maize field, exploitative competitive interactions will have an impact on the relative dominance of the different species, even if interference competition is low.

Distribution Range
There are several biological and behavioral characteristics that could either improve or impair the competitiveness of these pest species during both direct and indirect interactions (Table 1). First is the range of optimal climatic conditions needed for development, with temperature being the most important (Van Rensburg et al. 1985, Honek and Kocourek 1990, Denno et al. 1995, Ntiri et al. 2016, Yonow 2017, Babasaheb et al. 2018. Temperature affects nearly all aspects of the growth and development of insects, and most insects have adapted to function optimally only within a specific temperature range (Khadioli et al. 2014, Terblanche et al. 2015, Glatz et al. 2017, Garcia et al. 2018. The broader the range of temperatures that a species can tolerate, the less it will be constrained in its geographical distribution (i.e., the more agroecological zones it could inhabit- Fig. 4) and the greater its resilience would be to extreme climate conditions (Bale et al. 2002, Battisti and Larsson 2015, Yonow 2017, Azrag et al. 2018). Both C. partellus and S. frugiperda have higher optimal temperature ranges than B. fusca (Table 1). The composition of stem borer communities has been reported to change along an altitudinal gradient. Busseola fusca tends to dominate in cooler, higher altitudes, whereas C. partellus dominate in warmer lowland regions (Overholt et al. 2001, Zhou et al. 2003, Mwalusepo et al. 2015, Ntiri et al. 2019). In the midrange altitudes, the dominant species is determined by factors other than temperature. For instance, Van Rensburg et al. (1987) found that high levels of humidity (93%) resulted in significantly higher survival rates for B. fusca adults when compared with low levels of humidity (38% humidity) over a 4-d period. Consistent with the results for B. fusca, Tamiru et al. (2012) reported that C. partellus adult longevity  increased at higher humidity (80% vs 40%). However, C. partellus adult females displayed more resilience to changes in humidity, particularly when combined with high temperatures. No significant difference in adult female longevity was observed between the relative humidity levels of 40, 60, and 80% at 30°C (Tamiru et al. 2012). Busseola fusca populations might, therefore, be more dependent on rainfall (and the accompanying humidity) than C. partellus. With their comparable optimal temperature requirements, S. frugiperda will likely compete with C. partellus for dominance at the lower and midrange altitudes . Mwalusepo et al. (2015Mwalusepo et al. ( , 2018, Ntiri et al. (2019), and Mutamiswa et al. (2017a, b) discussed the possibility that climate change could have a significant effect on the structure of maize stem borer communities. The rising temperatures predicted for much of the African continent (Niang 2014) could detrimentally affect species such as B. fusca and S. calamistis and lead to the displacement of these once-dominant native species by invasive species (Mwalusepo et al. 2015, Mutamiswa et al. 2017a). The effect of climate change on the geographic distribution of insect pest species is a key concern in all agricultural regions across the globe (Björkman and Niemelä 2015). The altering of climatic conditions could cause insect pest species to invade traditionally cooler agroecological zones that were previously unsuitable for their development, allowing range expansion of pests toward the poles and toward higher elevations (Bale et al. 2002, Forrest 2016, Reineke and Thiery 2016, Castex et al. 2018. Concurrently, rising temperatures in the warmer tropics could lead to range contraction, since temperatures could exceed the upper thermal threshold for the development of insect pest species (Gutierrez et al. 2012, Reineke and Thiery 2016, Castex et al. 2018, Deutsch et al. 2018, Lehmann et al. 2020). Battisti and Larsson (2015) compiled a summary of the range expansions (ongoing or predicted) of insect pests in agriculture and reported 50 insect pest species of which the range either have been affected by climate change or are predicted to be affected in the future. Included in this list were important lepidopteran maize pests such as Ostrinia nubilalis (Hübner) (Crambidae) and Helicoverpa zea (Boddie) (Noctuidae). The range expansion of the latter two species into warming climates in North America was also forecasted by Diffenbaugh et al. (2008). The potential range expansion of H. zea was much greater than for the other species included in their study, and that this posed a risk to crops other than maize, since H. zea is a highly polyphagous migratory pest. They concluded that range expansions of pests could lead to severe economic impacts for the newly invaded agricultural systems.

General Development Time and Voltinism
Another biological trait which affects the indirect competitiveness of species (and which is closely linked to optimal temperature range) is generation time (from egg hatch to adult eclosion) and voltinism (Honek and Kocourek 1990, Bale et al. 2002, Altermatt 2010, Mwalusepo et al. 2015, Azrag et al. 2018. Generation time is closely linked with temperature and, although development rate generally increases with a rise in temperature, it is the fastest at the optimal temperature for the particular species (Honek and Kocourek 1990, Bale et al. 2002, Mwalusepo et al. 2015, Azrag et al. 2018, Babasaheb et al. 2018. Longer life cycles are disadvantageous to pest species since it limits its voltinism and the rate at which the population grows. Thus, the shorter the life cycle of a species, the greater the rate of population growth, and the more dominant the species will be as it relates to exploitative competition with other pest species (Kfir 1997, Reitz and Trumble 2002, Mutamiswa et al. 2017a. Busseola fusca has the longest life cycle of the three pest species, with an average generation time of 65.4 d at the optimal temperature of 25°C (Table 1). The effect of climate on stem borer moth flight patterns is illustrated by the differences between B. fusca moth flight patterns in the warm western region and cooler eastern Highveld region in South Africa. Development time of B. fusca is slower in the cooler areas, which limits it to two generations per cropping season, since they generally enter diapause for up to 5 mo during winter (Van Rensburg et al. 1985, 1987Kruger et al. 2012b;Calatayud et al. 2014;Glatz et al. 2017).
The life cycle of C. partellus is much shorter than that of B. fusca, but it commonly takes 10 to 12 d longer than that of S. frugiperda (Table 1). Chilo partellus can produce up to six generations per year (Table 1) (Bate et al. 1990, Mutamiswa et al. 2017a, and under ideal tropical conditions S. frugiperda may have up to eight generations per year (Luginbill 1928, Andrews 1988. Another testament to the advantage that climate change will confer to the invasive lepidopteran pest species is the fact that the duration of life cycles of both S. frugiperda and C. partellus decreases with an increase in temperature, while that of B. fusca is prolonged even further at increased temperatures (Table 1).
Climate change impacts (predominantly the increase of temperatures) on the voltinism of insect pests has been investigated by several authors (Porter et al. 1991, Bale et al. 2002, Steinbauer et al. 2004, Tobin et al. 2008, Jönsson et al. 2009, Altermatt 2010, Prasad et al. 2012, Kroschel et al. 2013, Fand et al. 2014, Khadioli et al. 2014, Fand et al. 2015, Babasaheb et al. 2018. Altermatt (2010) proposed that, apart from rapid larval development, an increased number of generations of lepidopterans per year could also be explained by an earlier onset of the flight period due to warmer temperatures in spring, and a change from univoltine populations to bi-or multivoltine populations that migrate from warmer agroecological zones to cooler zones as the growth season progresses. Babasaheb et al. (2018) reported on three lepidopteran pests for which an additional one to three generations per year are predicted by 2050: C. partellus (Khadioli et al. 2014), Phthorimaea operculella (Kroschel et al. 2013), and Spodoptera litura (Fand et al. 2015). Gagnon et al. (2019) also found that O. nubilalis populations in North America will likely develop an additional generation and have higher densities and wider distributions due to climate change, which will increase the pest status of this species in sweet corn.
Climate change could increase the impact of insect pests on crop yield (Bale et al. 2002, Björkman and Niemelä 2015, Babasaheb et al. 2018, since higher temperatures not only shortens development time of most species but also elevates their metabolism and population growth rates, which consequently increases their rate of consumption. A study conducted by Deutsch et al. (2018) found that a 2°C increase in the average global surface temperature would lead to a median increase of 31% of maize yield losses to insect pests globally. This translates to a loss of 62 metric megatons maize per year. It is expected that crop yields in temperate regions will be impacted more by insect pests than in tropical regions. This is due to the fact that warming in temperate regions will lead to increases in both the size of pest populations and their associated rates of consumption. In tropical regions, further warming could result in temperatures exceeding the upper limits of optimal conditions for insect pest populations (Deutsch et al. 2018, Lehmann et al. 2020. In many of the more temperate agroecological zones in Africa, maize yield losses to insect pests are expected to increase between 20 and 40% should the global surface temperature rise with 2°C (Deutsch et al. 2018).

Diapause and Off-Season Survival Strategies
Because insects are poikilothermic and, therefore, acutely affected by changes in the ambient temperature, they have evolved several strategies to survive suboptimal conditions. Diapause is a physiological example of such a survival strategy (Bale and Hayward 2010, Fand et al. 2012, Gill et al. 2017. Throughout Africa, B. fusca enters a facultative diapause during winter months (Usua 1970, Van Rensburg et al. 1987, Okuda 1990, Ebenebe et al. 1999, Calatayud et al. 2014. Chilo partellus can also enter a rest phase to escape unfavorable conditions (Kfir 1991, Kfir 1993), but S. frugiperda does not have this ability (Johnson 1987, Goergen et al. 2016, Harrison et al. 2019). This means that S. frugiperda populations cannot overwinter in temperate regions, that they invade during the warmer months and they have to recolonize regions when the environmental conditions become favorable during the following summer season , Early et al. 2018. The temporary absence of S. frugiperda from certain agroecological zones during winter months provides the two stem borer species with a competitive advantage during the off-season, since they are able to overwinter locally after which moths emerge and infest maize plants earlier during spring. However, C. partellus populations generally terminate their rest phase quicker than B. fusca does (~ 10 d vs 34 d) (Table 1). Consequently, where mixed populations occur, C. partellus populations are usually the first to infest maize plants, enabling the first generation to utilize the food and space resources without competition (Kfir 1997, Dejen et al. 2014, Mutamiswa et al. 2017a. Effectively, C. partellus populations are then able to outcompete the lagging B. fusca and S. frugiperda populations. In agroecological zones, which S. frugiperda have to invade seasonally, i.e., temperate highlands , Early et al. 2018, it does not have such a temporal competitive advantage.
Since the onset of diapause is primarily governed by temperature, humidity, and photoperiod, a warming climate may affect the life cycles of insects that undergo a winter diapause Hayward 2010, Gill et al. 2017). Warmer seasons might delay the initiation of diapause and cause the early termination thereof, which could result in low mortality in overwintering populations. The large early generations of pests and early crop colonizations would cause significantly greater damage to crops (Harrington et al. 2001, Sharma et al. 2005, Bale and Hayward 2010, Babasaheb et al. 2018. However, should the environmental cues for diapause induction become mismatched with the development of the insect population, the insect population will need to adapt by adjusting their responses to the cues for diapause behavior or face detrimental consequences (Forrest 2016). For example, another lepidopteran species, Lasiommata megera (L.) (Lepidoptera: Nymphalidae) (Van Dyck et al. 2015), have been predicted to suffer a reduction in population size due to attempting an additional generation in regions where summer temperatures are higher, but seasons are not yet long enough to allow completion of the second generation (Chinellato et al. 2014, Van Dyck et al. 2015, Forrest 2016). This has been the fate of O. nubilalis populations in Canada, since high temperatures and long day lengths during critical developmental stages are not conducive for diapause (Gagnon et al. 2019). On the other hand, the existence of genetic variability between uni-and bivoltine races of this species could, however, improve the ability of O. nubilalis to successfully adapt to the changing climate (Gagnon et al. 2019). Such changes could also occur in the case of B. fusca in highland regions where it currently has only two generations per cropping season, and where additional generations have been reported due to improved environmental conditions associated with cyclical changes in climate (Van Rensburg et al. 1985, 1987. In warmer African regions where the climate allows continuous planting of maize throughout most of the year, or where there are, e.g., two cropping seasons, host plants of varying growth stages are available for maize pests to infest. This continuous availability of maize plants results in less distinct moth flights of stem borers as well as overlapping generations in a single crop.

Larval Population Density
Another biological and behavioral trait that may enhance or reduce the competitiveness of a pest species is how it responds to high larval population densities. Decreased development time, lower larval mass, and increased mortality were observed in several studies on the effect of population density on intra-and interspecific competition in insect species (Goulson and Cory 1995, Gibbs et al. 2004, Fantinou et al. 2008, Underwood 2010, Flockhart et al. 2012, Ntiri et al. 2017, Himuro et al. 2018, Pavlushin et al. 2018. Population density affects the survival, development time, dispersal, choice of oviposition site, and adult size of lepidopteran pests (Van Hamburg 1980, Berger 1992, Fantinou et al. 2008, Flockhart et al. 2012, Ntiri et al. 2017. In a mixed population scenario, the development of a tolerant species may not be as severely affected by high larval densities than that of other less tolerant species in the population. This may confer to the former, significant size advantages in both direct and indirect competitive interactions (Delong et al. 2014). Ntiri et al. (2017) investigated the effect of larval density on the fitness and competitiveness of C. partellus, B. fusca, and S. calamistis and found that the intensity of competitive interactions was density dependant and that more intense inter-and intraspecific competition was displayed at higher densities. In their experiments, Ntiri et al. (2017) observed that, although the fitness of all three species of stem borers declined with increased larval density, C. partellus larvae were superior in interspecific interactions, since they displayed higher survival and relative growth rates than either B. fusca or S. calamistis when developing at 25°C. In another study, Ntiri et al. (2016) demonstrated that the dominance of a species during interspecific interaction was also dependant on temperature, since B. fusca and S. calamistis both displayed higher survival rates at low temperatures, compared to C. partellus, which was dominant at higher temperatures. Sokame et al. (2020) used a similar protocol as Ntiri et al. (2016) to study the interspecific competition among S. frugiperda and three stem borer species (B. fusca, C. partellus, and S. calamistis) under various constant temperature treatments. Results showed that C. partellus had the highest relative growth rate at nearly all temperature treatments and that it outcompeted not only the noctuid stem borers but also the noctuid leaf feeder (S. frugiperda). However, Sokame et al. (2020) also noted that the direct competition among S. frugiperda and the stem borers were limited in their study, since the stem borers entered the diet for feeding, whereas S. frugiperda larvae remained on the surface of the diet. The cannibalistic nature of S. frugiperda larvae is also thought to have limited the survival and possibly the growth rate of this species in the study. Although the laboratory study of Sokame et al. (2020) suggested that C. partellus would outcompete S. frugiperda at most temperature regimes, field studies with whole maize plants could lead to different outcomes.

Cannibalism and Predatory Behavior
Spodoptera frugiperda larvae respond to intra-specific competition at high larval densities with cannibalism (Luginbill 1928, Chapman et al. 1999). This behavior generally commences during the 3rd instar and is considered one of the main reasons why very low numbers of fully grown S. frugiperda larvae are found per plant, despite the high initial numbers of neonates . Although cannibalism comes with significant costs to fitness (Chapman et al. 1999), Downloaded from https://academic.oup.com/jipm/article-abstract/11/1/16/5875594 by guest on 29 July 2020 it also provides protection from predation via a decrease in the local larval density (Chapman et al. 2000). Additionally, the decrease in larval density also reduces the intra-specific competition for resources. However, S. frugiperda larvae are not merely cannibalistic (attack and devour conspecific larvae only), but they also engage in competitive intraguild interactions, where they attack the larvae of other species sharing the same resource (Luginbill 1928;Bentivenha et al. 2016Bentivenha et al. , 2017. Studies of intraguild interactions between S. frugiperda and Helicoverpa spp. (which also engages in cannibalistic behavior) showed that S. frugiperda had a greater survival when the larvae were the same instar or when S. frugiperda was larger than the competitor (Bentivenha et al. 2017). Thus, S. frugiperda had a competitive advantage over Helicoverpa spp., even though H. zea larvae displayed more aggressive behavior than S. frugiperda larvae (Bentivenha et al. 2016). It is, therefore, probable that S. frugiperda larvae would engage in direct interference competition behavior and attack the larvae of the two stem borer species when the larvae of these species are still feeding in the whorls of plants, prior to them boring into maize stems. Due to the rapid developmental rate of S. frugiperda (Table 1), its larvae will often have a size advantage over stem borer larvae that generally feed in the whorl only up to the 3rd or 4th instar (Van Rensburg et al. 1987, Calatayud et al. 2014, after which they commence boring into stems, whereas S. frugiperda larvae feed in the whorl until pupation.

Species Competition Summary
In this first part of the discussion, the biology and behavioral characteristics of the three pest species in question were compared in an attempt to uncover which species would dominate under specific conditions. The following generalizations can be made: • Busseola fusca will most likely be displaced by both C. partellus and S. frugiperda in the mild to warmer agroecological regions, but it will dominate in cooler areas due to its lower optimal temperature range and ability to diapause. However, this scenario may soon change due to the warmer and drier climate predicted for most of Africa. • Spodoptera frugiperda has a much shorter life cycle, and is aggressive and dominating in direct competition interactions. This may result in it outcompeting C. partellus in lowland humid agroecological zones where the latter was always dominant. However, the lack of diapause in the life cycle of S. frugiperda will benefit C. partellus in high-altitude areas where winter temperatures do not allow their survival.

Validity of IRM Assumptions for B. fusca, C. partellus, and S. frugiperda
The biology and behavioral characteristics of B. fusca, C. partellus and S. frugiperda, that influence the evolution of resistance in an HDR strategy are discussed in this section. The HDR strategy has proven to be a powerful tool to delay resistance evolution in target pest species if the underlying assumptions are valid (Tabashnik and Carrière et al. 2017) ( Table 2). The probable validity of the various assumptions regarding the three species of interest is compared below to assess the risk for resistance evolution of each species.

Initial Resistance Allele Frequency, Fitness Cost, and Incomplete Resistance
One of the five key assumptions for the HDR strategy (Table 2) to be effective is that the initial resistance allele frequency in the target population is low. It has often been suggested that a resistance allele frequency of <0.001 is required to ensure success of the HDR strategy (Gould 1998, Huang et al. 2011, Camargo et al. 2018). This assumption is suitable as a broad principle but could be too conservative in many cases. Tabashnik et al. (2013) suggested that factors such as the fitness costs associated with resistance could impact the minimum initial resistance allele frequency required for a population. For example, recessive inheritance of resistance together with fitness costs would still significantly delay resistance evolution even at an initial resistant allele frequency as high as 0.3, if refuges are present Tabashnik 2001, Siegfried andHellmich 2012). The initial frequency of resistant alleles most likely differs among populations and geographical areas, and, since no information exists on the status of susceptibility of these pests to Bt maize in Africa, it is not possible to validate this assumption for these species. Strydom et al. (2018), however, indicated that resistant alleles of B. fusca were already present in B. fusca populations in South Africa when initial efficacy evaluations of Cry1Ab maize were done during the mid-1990s. Botha et al. (2019) showed that although Cry1Ab maize provided field-level control of S. frugiperda in South Africa, larval survival on this single-gene Bt maize under laboratory conditions was as high as 35%. Vorster (2018) showed that C. partellus was still highly susceptible to both Cry1Ab maize and Cry1.105+ Cry2Ab2 maize. The important effect of high resistant allele frequencies in a population on the rate of resistance evolution was highlighted by Tabashnik et al. (2000Tabashnik et al. ( , 2005 who showed that unexpectedly high frequencies of resistance alleles were initially present in the populations of Pectinophora gossypiella (Saunders) (Lepidoptera: Gelechiidae) in Arizona in 1997.
This lack of information on resistant allele frequencies is also true for the assumptions about incomplete resistance and fitness costs associated with resistance (Table 2). Both fitness costs and incomplete resistance refers to trade-offs that occur during the evolution of resistance to Bt toxins (García et al. 2015, Paolino and Gassmann 2017, USEPA 2018. Incomplete resistance is indicated by reduced fitness of resistant individuals on Bt plants, whereas fitness costs are characterized by the greater fitness of susceptible individuals compared to resistant individuals on non-Bt plants (Tabashnik andCarrière 2017, USEPA 2018). These parameters could be key to delaying resistance evolution in pest populations that are under selection pressure from Bt maize.
A study of B. fusca populations resistant to Cry1Ab (MON810), which was conducted 6 yr after the initial report of resistance, could not detect fitness costs when these larvae were reared on Bt maize (Kruger et al. 2014). This suggests that resistance became a stable trait in these B. fusca populations. For S. frugiperda, several studies have indicated the presence of fitness costs to Cry1F, Vip3A and a pyramid maize event (which incorporated Cry1F, Cry1.105, and Cry2Ab) in populations from the Americas (Jakka et al. 2014;Dangal and Huang 2015;Horikoshi et al. 2015;Bernardi et al. 2016Bernardi et al. , 2017. However, the USEPA (2018) indicated that the fitness costs associated with Cry1F were not observed in any field-collected strains of S. frugiperda in the United States. Jakka et al. (2014) also found that the lower rate of larval development which was flagged as a fitness cost in Cry1F-resistant S. frugiperda populations in Puerto Rico did not affect the stability of resistance after 12 generations of selection of a heterogeneous S. frugiperda strain raised on a meridic diet. It would, therefore, be reasonable to assume that B. fusca and S. frugiperda populations present in the other regions in Africa could exhibit a similar lack of fitness costs, incomplete resistance and high frequency of resistant alleles to single-toxin maize events, which will necessitate the use of multi-toxin events from the outset.

Recessive Inheritance of Resistance (High-Dose Requirement of Toxin Expression)
Another assumption to be considered is also related to the genetics and heritability of the resistance trait. The HDR strategy assumes that the inheritance for resistance is recessive. This means that a heterozygous individual will not be able to survive on Bt plants, which then contributes largely to delay resistance evolution in a population (Tabashnik et al. 2013, Campagne et al. 2016). This assumption is alternatively expressed in terms of the high-dose requirement for a Bt crop since the toxin expression within Bt plants must ideally be high enough to kill all heterozygous individuals (Tabashnik and Carrière 2017). Theoretically, a high dose should constitute 25 times the dose needed to kill 99% of susceptible pest individuals (Gould 1998, USEPA 1998, Glaser and Matten 2003, Campagne et al. 2013. The lack of a high-dose expression of Cry1Ab maize was considered a contributing factor to the rapid evolution of resistance in the South African B. fusca populations (Van Rensburg 1999; Fitt et al. 2004;Tabashnik et al. 2008Tabashnik et al. , 2009Campagne et al. 2013;Strydom et al. 2018). Similarly, both Cry1F and Cry1Ab was shown not to meet the high-dose requirements for S. frugiperda and have since been rendered obsolete in Brazil due to resistance development by the target pest populations (Farias et al. 2016, Botha et al. 2019. In fact, Tabashnik and Carrière (2017) indicated that, of the 16 cases of practical resistance of insect pests to Bt toxins, not one of the Bt toxins achieved the high-dose standard for the target pest species.
Interestingly, no cases of field-evolved resistance to Bt maize of event MON810 (or any other event) have been observed for C. partellus or S. calamistis, even though these species have been subjected to the same selection pressure as B. fusca populations over the two decades that this maize event has been cultivated in South Africa. Similarly, Tabashnik and Carrière (2019) reported on 19 cases where no significant decrease in susceptibility to Bt crops have been recorded after a mean of 10 yr of exposure to the Bt toxin. The lack of resistance evolution in C. partellus and S. calamistis may indicate that event MON810 is a high-dose event for these species, that the frequency of resistant alleles is very low, and that fitness costs are associated with resistance evolution in the latter two species. Van Wyk and Van den Berg (2008), Van den Berg and Van Wyk (2007) reported that S. calamistis was highly susceptible to Cry1Ab maize and Van Rensburg (1999) reported similarly for C. partellus.

Multi-Toxin Events (Pyramids)
Considering the history of resistance evolution of B. fusca and S. frugiperda against single-toxin maize events (Cry1Ab and Cry1F), it would be sensible to immediately employ a pyramided Bt maize event to proactively address invalid IRM assumptions for populations of these two pest species in Africa (Head and Greenplate 2012, Storer et al. 2012. Important to note, however, is that the use of multi-toxin crops brings with it an additional three assumptions that must be satisfied to ensure the sustainability of a pyramid event (Table 2)  . For the purpose of this discussion, we focus on a pyramid event, MON89034, which expresses Cry1.105 and Cry2Ab2 proteins. Maize plants with this event are cultivated in South Africa and have been shown to be effective against both B. fusca and S. frugiperda populations (Strydom et al. 2018, Botha et al. 2019. The first of the pyramid-specific assumptions related to IRM is that each toxin in the pyramid should meet the high-dose standard for the target pest. MON89034 expresses Cry1.105 and Cry2Ab2 toxin proteins. Cry1.105 is a chimeric gene, containing domains I and II of Cry1Ab and Cry1Ac, respectively, and domain III of Cry1F Hellmich 2012, USEPA 2018). Recent findings confirmed that cross-resistance is highly likely between toxins with similar amino acid sequences in domain II (Hernandez-Rodriguez et al. 2013, Welch et al. 2015. This could compromise the efficacy of Cry1.105 against both S. frugiperda and B. fusca and indicates that this protein does not meet the high-dose requirements for both these target species. Although Strydom et al. (2018) confirmed that MON89034 is Downloaded from https://academic.oup.com/jipm/article-abstract/11/1/16/5875594 by guest on 29 July 2020 still highly effective against B. fusca, they highlighted the inherent low sensitivity of this species to Cry proteins. Low sensitivity combined with the low-dose expression of Cry1Ab in MON810 maize create an environment for resistance evolution of B. fusca to pyramid maize in South Africa. Similarly, Botha et al. (2019) found only moderate sensitivity of S. frugiperda to Cry1Ab, which they ascribed to MON810 being low-dose event for this pest, and to the possibility that the population that has invaded Africa could already have contained alleles with resistance to this protein. Because Cry1 and Cry2 proteins do not share binding sites (Hernandez-Rodriguez et al. 2013), it is safe to assume that MON89034 fulfills the requirement of the second assumption for pyramided crops, no cross-resistance (Table 2).
Finally, an important IRM requirement in farming systems where pyramid events are cultivated, is that multi-toxin plants are not grown simultaneously with single-toxin plants that produce one of the toxins expressed in plants of the pyramid event (Table 2). This is because the single-toxin event would compromise the efficacy of using two toxins against a pest population since it would allow the stepwise evolution of resistance against the pyramid variety (Zhao et al. 2003, Bates 2005, USEPA 2018. This assumption will conceivably only be violated in South Africa ( Van den Berg et al. 2013, Strydom et al. 2018, if other African countries decide to completely forego the first generation single-toxin events, and only commercialize pyramid events. In spite of evidence that the cultivation of MON810 maize alongside MON89034 could compromise the efficacy of the pyramid event, this has been ongoing for several years in South Africa. This is in contrast to countries such as Australia, where the simultaneous cultivation of both single-toxin Cry1Ac and pyramid Cry1Ac + Cry2Ab cotton has been prohibited, and where the replacement of the single toxin event with the pyramid was accomplished in a single year (Downes andMahon 2012, Carrière et al. 2016).

Abundant Refuges of Non-Bt Host Plants
Any plant species in a particular agroecosystem can serve as a refuge plant, provided it is a suitable host of the target species. Abundant and appropriate non-Bt refuges near Bt maize fields are critical components of the HDR strategy. Without these refuges, the selection pressure for the evolution of resistance would be severe, since very few susceptible alleles would remain in the population after even a single exposure to the Bt crop (Gould et al. 1998, USEPA 1998, Bates et al. 2005. The lack of sufficient refuges has contributed to the evolution of resistance in several pest populations, including B. fusca in South Africa (Van den Berg et al. 2013), P. gossypiella in India (Mohan et al. 2016, Tabashnik andCarrière 2019), S. frugiperda in Brazil (Monnerat et al. 2015), and Diabrotica virgifera virgifera (LeConte) (Coleoptera: Chrysomelidae) in the United States (Tabashnik and Gould 2012). The planting of separate, dedicated areas of non-Bt maize near Bt maize fields is considered the best option for IRM in most situations, but not all (Carroll et al. 2012). Smallholder farms are the norm in Africa, and mandating the planting of a separate refuge by each farmer that cultivates Bt maize would be impractical (Bates et al. 2005, Carroll et al. 2012, Li et al. 2017. It has been suggested that if the adoption rates for Bt maize in these farming systems are low, that the non-Bt maize fields of non-adopters in the area would suffice as refugia (Van den Berg and Campagne 2015). It is also conceivable that several farmers could plant a communal refuge area to service all their Bt fields (Head 2004). Still, the management of these refuges would require training of farmers and ongoing support from government or industry extension personnel, both services that are generally inefficient or lacking in smallholder agricultural areas (Jacobson and Myhr 2012, Kotey et al. 2016, Kotey et al. 2017.

Naturally Occurring Refuges
The suggestion has been made that the implementation of structured refuges might not be necessary for smallholder farmers if sufficient wild host plants are present in the uncultivated areas near smallholder plots (Head 2004, Onstad and Carrière 2014, Li et al. 2017. While investigating this claim for maize stem borers in Africa, Van den Berg (2017) compiled a list of conditions that must be met if wild host plants are to be used as a refuge in an HDR strategy (Table 2), and concluded that wild host plants are unsuitable for use as refuges for African maize stem borers. Instead of serving as refuges for the survival and production of susceptible individuals of the target pest species, wild grasses were found to fulfill the role of dead-end-trap plants. Wild host plants were found to be less attractive to the stem borers as oviposition sites and did not allow a high enough development or survival rate of the larvae, which resulted in the emergence of a limited number of adults that could compete for mating opportunities with resistant individuals (Van den Berg 2017, Li et al. 2017). Consequently, even though both B. fusca and C. partellus are able to survive and propagate to a very limited degree in these uncultivated habitats (C. partellus more so than B. fusca) (Ofomata et al. 2000, Songa et al. 2002, Mutamiswa et al. 2017a, unstructured refuges in the form of wild host plants will not be an IRM option for these species (Van den Berg 2017, Li et al. 2017, Sokame et al. 2019a.
Unlike the oligophagous nature of the two stem borer species, S. frugiperda is highly polyphagous, with approximately 350 host plant species (Montezano et al. 2018). It is primarily a pest of maize, but also shows preference for other poaceous plants, including crops such as millet (Panicum miliaceum), sorghum and rice (Oryza sativa). However, feeding damage has been observed on cowpea (Vigna unguiculata), groundnut (Arachis hypogaea), potato (Solanum tuberosum), soybean (Glycine max), and cotton (Gossypium hirsutum) (Goergen et al. 2016, Day et al. 2017, Assefa and Ayalew 2019. The richness of alternative host plants in diverse agricultural landscapes (characteristic of smallholder farming systems in Africa) (Van den  could mean that the implementation of structured refuges for S. frugiperda would not be necessary if Bt maize is deployed for control of this pest (Head 2004, Onstad andCarrière 2014). Although, this would necessitate that the abundance and reliability of alternative hosts are determined before the approval of Bt maize and would also need to be monitored to ensure their constant availability (Head 2004, Van den Berg 2017. At this stage, however, the polyphagous nature of S. frugiperda has not materialized in Africa, and host plants other than maize that are attacked are sorghum and millet. This limited host association is ascribed to S. frugiperda in Africa being an inter-strain hybrid with a limited affinity for crops other than maize and sorghum (Nagoshi 2019. In a case similar to S. frugiperda, the possibility of using other crops as refuges for Bt cotton seems viable for a highly polyphagous pest such as Helicoverpa armigera (Hübner) (Lepidoptera: Noctuidae). A study of the population dynamics of H. armigera in China showed that other crops can be used as a refuge for H. armigera in Bt cotton-growing areas (Ye et al. 2015). This was also demonstrated with Heliothis spp. infesting cotton in the United States, or other crop species refuges such as tobacco for Bt cotton in the United States, and pigeon pea for Bt cotton in Australia (Andow 2008). Indigenous plants and weedy hosts of H. armigera were also indicated to provide appropriate refuges in Bt cotton production areas in South Africa (Green et al. 2003). The legume pod borer, Maruca vitrata (Fabricius) (Lepidoptera: Crambidae), the target pest of Bt cowpea in West Africa, is a polyphagous pest of grain legumes that has a wide distribution throughout tropical and subtropical regions worldwide (Agunbiade et al. 2014). The populations of M. vitrata that are sustained on alternative host plants can contribute to IRM by sustaining individuals that have not been subjected to selection pressure on Bt cowpea (Onstad et al. 2012).

Seed Mixtures for IRM in Smallholder Farming Systems
An alternative approach to structured refuges is the use of seed mixtures. This approach has been identified as a possible solution to the challenges facing IRM for Bt maize in smallholder farming systems since it does not require the farmer to plant a separate structured refuge (Carroll et al. 2012, IRAC 2013, Erasmus et al. 2016, Carzoli et al. 2018. Not only will this approach be more practical to implement at smaller scales of production, but it also ensures farmer compliance to refuge requirements, since the seed is premixed (Head 2004, Head andGreenplate 2012). The spatial integration of Bt maize and refuge areas also improves the likelihood that resistant and susceptible adult individuals will mate, compared to the separate structured refuges (Onstad et al. 2011, Spencer et al. 2013. Seed mixtures could also prevent gravid female moths from preferentially ovipositing on Bt plants. This phenomenon was observed by Téllez-Rodríguez et al. (2014) with S. frugiperda in Cuba. The preference for Bt maize was most likely due to damage avoidance behavior (favoring undamaged Bt plants over the heavily damaged refuge plants), and not due to an outright preference for Bt maize. However, in a recent study, Gonçalves et al. (2020) found that S. frugiperda females did not distinguish between Bt and non-Bt plants when selecting an oviposition site, and that conspecific larval damage only affected oviposition choice in a minor capacity. Nascimento et al. (2020) also indicated highly variable oviposition responses by female moths to Bt and non-Bt maize, with preference for Bt maize in some cases. However, they also indicated that preference for oviposition on a particular hybrid was influenced by whether plants were already infested and damaged by larvae, as well as by physical characteristics of leaves, e.g., hairiness. Females of lepidopteran pests generally do not show any oviposition preference for either Bt or non-Bt plants that are of the same age and quality Rice 2001, Visser et al. 2019). Additionally, several studies have found no significant differences in the volatile profile of Bt and non-Bt maize and rice plants when the degree of damage to the plants were controlled (Dean and De Moraes 2006, Sun et al. 2013, Liu et al. 2015. Although Turlings et al. (2005) and Xu et al. (2019) found differences in the amount of specific volatiles released by Bt and non-Bt maize, the differences in the volatile profiles of plants from the same species that arise from conventional breeding programs are far greater than the changes caused by genetic engineering (Turlings et al. 2005;Vogler et al. 2009Vogler et al. , 2010Xu et al. 2019), and therefore should not greatly impact the preference of gravid female moths between Bt and non-Bt plants.
The damage avoidance behavior observed by Téllez-Rodríguez et al. (2014) may increase the selection pressure exerted on the pest population, increasing the rate of resistance development (Téllez-Rodríguez et al. 2014, Visser et al. 2019. Seed mixtures could prevent the buildup of high pest population densities, thereby preventing the total decimation of refuge plants in the case of S. frugiperda. Contrary to the damage avoidance oviposition behavior observed for S. frugiperda, Sokame et al. (2019b) observed that adults of B. fusca, C. partellus, and S. calamistis showed positive responses to volatile organic compounds emitted by plants infested by larvae (either conspecific or heterospecific) over those from uninfested plants. These inclinations further translated to the preferential selection of oviposition sites on infested plants versus uninfested plants (a phenomenon also observed by Ntiri et al. (2018)). Using gas chromatography/ mass spectrometry and chemical analysis, Sokame et al. (2019b) discovered that the volatiles emitted from infested plants were compositionally richer than those released by uninfested plants.
The biggest threat to the sustainability of the seed mixture approach is the propensity for movement of pest larvae. Several studies have indicated that if the larvae of the target pest are prone to migrate between Bt and non-Bt plants, seed mixtures may lead to accelerated evolution of resistance (Mallet and Porter 1992, Roush 1997, Carrière et al. 2016. This would occur due to two processes: firstly, reduced survival of susceptible insects (and effectively the refuge size), and secondly, increased survival of heterozygotes, which leads to a rise in the frequency of resistant alleles in the population (Head and Greenplate 2012, Carrière et al. 2016, Onstad et al. 2018. It is, therefore, important to take into account the larval movement behavior of the target pest when considering the use of a seed mixture as IRM strategy. Pannuti et al. (2016) investigated the plant-toplant movement of S. frugiperda in non-Bt maize plots and found that 92% of S. frugiperda larvae were recovered within a 1.1 m radius from the inoculated plant, and more than half of these were found within the same row as the natal plant, 2 wk after inoculation. Given the inter-and intra-row spacing at 0.76 m and 0.15 m, respectively, this radius roughly translates to an average movement across seven to eight maize plants within the same row. Visser et al. (2020a) reported that, on average, B. fusca larvae were found no further than two plants away from the inoculated plant in a non-Bt maize field after 3 wk in the field (0.75 m inter-and 0.20 m intrarow spacing). Although few similar studies have been conducted for C. partellus on maize, it has been reported that the larvae of this species have a higher propensity for inter-plant movement than B. fusca, when these species were compared on sorghum (Van den Berg et al. 1991b). Chapman et al. (1983) found that, within a week after inoculation, C. partellus larvae migrated from a single inoculated sorghum plant to infest a total of seven sorghum plants in a stand with an inter-and intra-row spacing of 0.75 and 0.10 m, respectively. In a study of C. partellus larval dispersal on maize, Päts and Ekbom (1992) observed that the mean number of plants infested from a single egg batch after a 7-d period ranged between 2.3 to 4.2 (interrow spacing = 0.7 m, and intra-row spacing = 0.3 m). The higher level of larval movement observed by Chapman et al. (1983) compared to that of Päts and Ekbom (1992) can, in part, be attributed to the lower plant density used in the latter study. The low larval survival rate described for all the above-mentioned species was reported to be common for lepidopteran species under field conditions (Zalucki et al. 2002).
The propensity of all three species to migrate between host plants should, in theory, preclude the use of seed mixtures as IRM strategy (Heuberger et al. 2011, Carrière et al. 2016, Visser et al. 2020a. However, Visser et al. (2020a) suggested that, in practice, there are ways to mitigate the successful movement of the pest larvae. A decrease in plant density, for example, in an intercropping system would significantly affect the ability of larvae to successfully migrate between plants, since they are more prone to desiccation in heat and wind, drowning in rain and falling victim to predators and parasitoids (Bonhof 2000, Zalucki et al. 2002, Midega et al. 2006, Visser et al. 2020a). Intercropping with non-host plants, which is Downloaded from https://academic.oup.com/jipm/article-abstract/11/1/16/5875594 by guest on 29 July 2020 a common practice in African smallholder farming systems, could further contribute to decreased survival of migrating larvae since the presence of non-host crop plants would not only decrease the density of host plants and serve as obstructions to migrating larvae, but a higher plant diversity increases the presence of natural enemies within the maize field (Khan et al. 1997, Landis et al. 2000, Midega et al. 2006. Khan et al. (2014) and Midega et al. (2018) have also indicated the utility of push-pull systems in controlling both stem borer species and S. frugiperda. This technology uses the drought tolerant greenleaf desmodium, Desmodium intortum (Miller) (Leguminosae), and Brachiaria grass (cv Mulato II) (Poaceae) as the 'push' and'pull' crops, respectively (Khan et al. 2014, Midega et al. 2018). In their study, Midega et al. (2018) found that, compared to pure stands of maize, reductions of 82.7% in average number of S. frugiperda larvae per plant and 86.7% in plant damage per plot were observed in the push-pull stands.
The success of seed mixture strategies against ear-feeding lepidopteran species such as S. frugiperda could also be undermined by cross-pollination of Bt and non-Bt maize and the resulting formation of mosaic kernel tissue in maize ears. Field studies have established that cross-pollination is more likely when Bt and non-Bt maize are grown in close proximity (1-4 m) (Chilcutt and Tabashnik 2004, Burkness and Hutchison 2012, Hofmann et al. 2014). Yet, in a study evaluating a decade of monitoring data, Hofmann et al. (2014) found that substantial amounts of maize pollen are distributed via wind over distances in excess of 100 m (and even further than 1,000 m), with deposition values of 3,000 to 164,000 maize pollen grains per square meter. The use of separate refuges to reduce the prevalence of mosaic kernel tissue is advisable in large-scale commercial farms where several hundred hectares of maize are planted. However, in small farming systems, it is unlikely that the use of separate refuges will significantly reduce the prevalence of cross-pollination, due to small field sizes and the close proximity of refugia to Bt fields.

IRM Assumptions Summary
Validity of HDR strategy assumptions for managing insect resistance evolution by C. partellus, B. fusca, and S. frugiperda to Bt maize in Africa: • Both B. fusca and S. frugiperda have high inherent risks of evolving resistance to single-toxin Bt maize due to preexisting high resistance allele frequencies within populations and a lack of fitness costs. After more than two decades of selection pressure on Bt maize in South Africa, no resistance has yet been detected for C. partellus or S. calamistis, suggesting that singletoxin events may be effective to manage these pests successfully throughout the continent. • The MON89034 event is high-dose for all three species, ensuring recessive inheritance of resistance. However, since one of the toxins produced in the pyramid is not high-dose for both B. fusca and S. frugiperda, care must be taken that this pyramid is not cultivated in tandem with single-toxin events of the weaker toxin. • Due to the oligophagous nature of B. fusca and C. partellus, and extremely low carrying capacity of their wild host plants, these hosts will not suffice as refuges in IRM programs, necessitating the planting of non-Bt maize refuges for both these species. The conservative feeding strategy of the inter-hybrid strain of S. frugiperda that occurs in Africa, also limits its effective host range, with only cultivated sorghum that could be considered as an alternative to non-Bt maize refuges.
• In spite of the propensity for larval movement between maize plants, seed mixtures could be a viable approach to refuge implementation for all three species when combined with IPM practices such as intercropping and push-pull systems in smallholder farming systems.

IRM Options for Single and Mixed Populations of Lepidoptera Maize Pests in Africa
The information provided in the sections above and the implications thereof on IRM strategies against these pests are discussed below in terms of the three options available for IRM in African smallholder farming systems. These options are structured refuges, seed blends, and/or combinations of these with IPM strategies that suppress pest populations. Firstly, irrespective of IRM strategy, it would be most advantageous for countries in Africa to immediately opt for the use of a pyramid event and forego the use of single-toxin events (e.g., MON810), especially in regions where the prevalence of mixed populations of the species in question are high. Single-toxin events should only be considered in cases where it is targeted against a pure population of C. partellus, which is not an option anymore with the invasion of S. frugiperda in Africa. We suggest that S. frugiperda will most likely dominate the Lepidoptera species complex in non-Bt maize fields in moderate to warm agroecological zones if it occurs in mixed populations with B. fusca and C. partellus.

Structured Refuges
Due to the limited sizes of fields in smallholder farming systems, only a very small area can be devoted to planting a separate refuge of non-Bt maize, if structured refuges are enforced. If such structured refuges are the recommended IRM strategy, inter-species interactions could influence refuge efficacy. The biology and behavior of S. frugiperda may enable it to outcompete the other two species, displacing them from the refuge plants. This would increase the selection pressure on stem borers since fewer adults will emerge from the refuge to mate with the resistant individuals that might survive on the Bt maize.
High population densities of S. frugiperda, coupled with its short development time, could also lead to the complete decimation of plants in small refuge areas. This would limit the available resources and compel gravid females of both C. partellus and S. frugiperda to preferentially oviposition on plants inside the Bt field, since the females of both these species oviposit on the leaves of maize plants. Busseola fusca might still oviposit in the highly damaged refuge area since gravid female moths show preference for damaged maize plants and does not lay egg son leaves but beneath leaf sheaths. However, the resulting B. fusca larvae would need to compete with the more dominant and aggressive S. frugiperda and C. partellus larvae that remain in the refuge and could suffer high mortality as a result. This would decrease the number of susceptible B. fusca individuals available to mate with resistant individuals. Spodoptera frugiperda populations could escape the increase in selection pressure on Bt maize if alternative host plants are available as refuge plants. However, alternative host plants will not suffice as refuge for the B. fusca and C. partellus populations. The use of separate refuges might, therefore, not be advisable in warmer agroecological zones where mixed populations of lepidopteran pests are common.
In more temperate climates, where winter temperatures prohibit the overwintering of S. frugiperda, the effect of this pest on plants grown in small refuges could be delayed until later in the summer season when S. frugiperda reestablishes itself in an area. In the case of C. partellus, on the other hand, its early termination of diapause, higher rate of development, and proclivity to migrate would result in it becoming the dominant species in the refuge area, possibly displacing B. fusca from the refuge in the case of high population densities. Since B. fusca is less susceptible to Bt toxins and one of the toxins within the MON89034 event does not meet high-dose requirements, evolution of resistance against the pyramided crop could be accelerated. Therefore, even in the absence of S. frugiperda, mixed seed refuges could be a more sustainable IRM strategy for mixed stem borer populations in smallholder farming systems

Seed Mixtures
This approach to IRM is easier and more practical to implement in smallholder farming systems and ensures farmer compliance to refuge requirements. However, pest behavior and larval movement within such systems may have a strong influence on its efficacy to delay resistance evolution. The spatial integration of Bt maize plants and refuge areas (non-Bt plants) also improves the likelihood that resistant and susceptible adult individuals will mate, compared to the separate structured refuges. Although the propensity of all three lepidopteran species to migrate between host plants theoretically precludes the use of seed mixtures, this could be addressed by means of existing farming practices which adversely affect pest populations.

IPM Strategies
The use of IPM strategies in lieu of the reliance on a single pest control tactic (e.g., Bt crops, or chemical insecticides) is the most sustainable approach to insect pest management Hutchison 2015, Anderson et al. 2019, Tabashnik and Carrière 2019. Farming practices that suppress pest populations by making the crop environment unfriendly for pests to colonize, establish, survive, and multiply, have long been used in Africa. For example, intercropping practices disrupt movement of stem borer larvae enhances mortality within maize fields (Khan et al. 1997, Landis et al. 2000, Midega et al. 2006. Trap cropping with Napier grass has also been shown to result in reduced pest numbers of several stem borers species in maize (Van den Berg and Van Hamburg 2015). The climate-adapted push-pull system which consists of maize intercropped with Greenleaf desmodium (Desmodium intortum), used as the push component, and Brachiaria grass (pull) planted around maize fields, serving as the pull component or trap crop has been reported to result in a 97% reduction in S. frugiperda numbers in maize in East Africa where this system has been adopted by smallholder farmers (Midega et al. 2018). The activity of predators and parasitoids has also been shown to be higher in such diverse cropping systems (Bonhof 2000, Midega et al. 2006. Baudron et al. (2019) showed that agronomic practices such as frequent weeding as well as minimum-and zero-tillage, effectively reduced S. frugiperda numbers in typical African smallholder conditions. Harrison et al. (2019) summarized numerous low-cost, smallholder friendly solutions for S. frugiperda control in Africa, many of which relate to agronomy and habitat management that reduces pest colonization and survival in maize fields.

Integration of Structured Refuges or Seed Mixtures Into IPM Systems
The use of seed mixtures together with farming practices that reduce colonization, establishment, and survival of these lepidopteran pests may be a viable strategy to manage resistance evolution of pests to Bt maize in Africa.
The implementation of most IPM practices is generally impractical at large scales of production, whereas several of these IPM measures (e.g., intercropping) are already in widespread use among smallholder farmers (Tefera et al. 2016). By combining seed mixtures with these habitat management practices, the risk that larval movement poses to resistance management can be mitigated (Midega et al. 2006).
The value of combining IPM and IRM strategies to limit resistance evolution, especially in developing countries where the implementation of the standard IRM practices are complicated by several factors, is illustrated by the case of P. gossypiella in India (Tabashnik and Carrière 2019). The extensive nature of the resistance of P. gossypiella to Bt cotton in India necessitates the return to IPM programs that also employ control tactics other than Bt toxins (Kranthi 2015, Mohan 2017, Tabashnik and Carrière 2019). For P. gossypiella, these approaches include planting of early to medium maturing cotton hybrids, removal of crop residues and ratoon cotton, use of insecticides, crop rotation, mass trapping and mating disruption, and biological control with natural enemies (Kranthi 2015, Mohan 2017, Tabashnik and Carrière 2019).

Summary of IRM Options for Lepidoptera Maize Pests in Africa
• Inter-species interactions could adversely influence the efficacy of structured refuges. A limitation of this strategy is that noncompliance will be common, and it may only be effective in certain agroecological zones where only one or two species of the pest complex occur. Furthermore, due to the limited sizes of fields in smallholder farming systems, only a very small area can be devoted to planting a separate refuge of non-Bt maize, if structured refuges are enforced. • Seed mixtures: Although theory indicates that larval movement between host plants limits the use of seed mixtures, this could be mitigated by means of existing farming practices, which adversely affects pest populations. This approach to IRM is more practical and can be implemented in smallholder farming systems, thereby also addressing the issue of noncompliance. • Integrated Pest Management: Selection pressure for resistance evolution may be reduced by disrupting pest biology and suppressing of pest numbers in maize-based agro-ecologies. This can be done through the use of different pest management strategies, particularly cultural control, which is already widely used among smallholder farmers. Seed mixture refugia may, therefore, be a solution to IRM challenges faced in smallholder farming systems in Africa. IRM will be more effective and more valuable when it is incorporated into integrated pest management (IPM). The best IRM will take advantage of the best IPM (Onstad 2014).

Conclusions
This review suggests that the use of structured separate refuges as components of IRM strategy against mixed pest populations in smallholder Bt maize fields would be unwise. A seed mixture approach, coupled with IPM practices, may be more sensible, since it may limit the opportunity for a single species to dominate the species complex and cause structured refuges to be unsuitable for other species, which may lead to increased infestation of Bt plants, and subsequent increased selection pressure. Furthermore, seed mixtures also address the issue of non-compliance.
This thought experiment was limited to the effect that scale of production, climatic conditions, and pest population dynamics could have on the choice and design of an IRM strategy for Bt maize in Africa. Future studies should aim to test the hypotheses laid out in this text by producing resistance evolution models for specific areas. These models should consider agroecological zones, as well as population dynamics of the pest complex, regarded the target of Bt maize in a particular region.