Effects of distance from semi-natural habitat on fall armyworm (Spodoptera frugiperda, J. E. Smith) and its potential natural enemies in Ghana

Abstract Insect crop pests are a major threat to food security in sub-Saharan Africa. Configuration of semi-natural habitat within agricultural landscapes has the potential to enhance biological pest control, helping to maintain yields whilst minimising the negative effects of pesticide use. Fall armyworm (Spodoptera frugiperda, J. E. Smith) is an increasingly important pest of maize in sub-Saharan Africa, with reports of yield loss between 12 and 45%. We investigated the patterns of fall armyworm leaf damage in maize crops in Ghana, and used pitfall traps and dummy caterpillars to assess the spatial distribution of potential fall armyworm predators. Crop damage from fall armyworm at our study sites increased significantly with distance from the field edge, by up to 4% per m. We found evidence that Araneae activity, richness and diversity correspondingly decreased with distance from semi-natural habitat, although Hymenoptera richness and diversity increased. Our preliminary findings suggest that modifying field configuration to increase the proximity of maize to semi-natural habitat may reduce fall armyworm damage and increase natural enemy activity within crops. Further research is required to determine the level of fall armyworm suppression achievable through natural enemies, and how effectively this could safeguard yields.


Introduction
The human population of sub-Saharan Africa is expected to at least double by 2050, necessitating a corresponding increase in food availability (Ittersum et al., 2016;Cleland and Machiyama, 2017). Yields per hectare have, however, remained largely static in recent decades, and many African countries are net importers of food (Cleland and Machiyama, 2017). Sustainable intensification to close the yield gap between current and maximum possible production could help improve food security in this region (Godfray et al., 2010;Godfray and Garnett, 2014).
One barrier to improving food security is yield loss from invertebrate pest damage. Sustainable intensification includes an increased reliance on biological pest control (Garibaldi et al., 2017). As part of integrated pest management, this could increase yields while minimising the negative impacts of pesticide application such as environmental pollution and compromised human health (Bianchi et al., 2006;Kansiime et al., 2019;Tambo et al., 2020;Haggblade et al., 2021). The configuration of landscape natural capital stocks (e.g. semi-natural habitat surrounding cropland) can potentially be managed to increase the flow of biological pest control services, particularly on smallholder farms (Steward et al., 2014).
One such biological pest control service could come from habitat spillover effects, where populations of invertebrates from surrounding semi-natural habitat move into crop areas to consume or parasitize pests (Tscharntke et al., 2005). Here, we refer to these predators and parasitoids as 'natural enemies'. Complex landscapes consistently contain higher natural enemy populations than simple landscapes (Bianchi et al., 2006). Furthermore, richness and diversity of natural enemy communities is important in delivering effective biological pest control (Wilby and Thomas, 2002;Cardinale et al., 2003;Crowder et al., 2010). However, measures to diversify crop landscapes are not guaranteed to enhance natural enemies, decrease pests, or increase yields (Bianchi et al., 2006;Poveda et al., 2008). Semi-natural habitat can act as reservoirs for crop pests, increasing their abundance and corresponding crop damage (Tscharntke et al., 2005). The inconsistent impact of semi-natural habitat limits the ability to make general recommendations for landscape management. Therefore, localized case-specific studies are required to inform management recommendations.
Fall armyworm (FAW, Spodoptera frugiperda J. E. Smith, Lepidoptera: Noctuidae), native to the Americas, has been invasive in West Africa since 2016 and attacks several economically important crops such as maize, millet, sorghum, rice, wheat, and sugar cane (Goergen et al., 2016). FAW is a particular problem on maize (Zea mays) (Day et al., 2017;Early et al., 2018); estimates of yield loss in sub-Saharan Africa range between 12 and 45%, with 26% losses reported by Ghanaian farmers in 2018 (Njuguna et al., 2021). Although a variety of natural enemies of FAW have been identified in Ghana (Agboyi et al., 2020;Koffi et al., 2020) and studies on landscape management for its control are emerging from sub-Saharan Africa (e.g. Midega et al., 2017), studies investigating how abundance of FAW or its natural enemies changes with distance from field edges are lacking (Harrison et al., 2019). This is fundamental to understanding the potential for semi-natural habitat to enhance biological control of this economically important pest, which in turn is required to inform management recommendations for landscape configuration to minimise pest abundance and maximise crop yield.
Using a combination of empirical observation and experimental dummy caterpillars in Ghanaian maize fields we examine the relationship between distance from semi-natural habitat and FAW-mediated crop damage, and the abundance, richness and diversity of its potential natural enemies. We hypothesise that natural enemies colonise maize from semi-natural habitat, and therefore predict that FAW crop damage will be lower, and the activity of natural enemies higher, closer to field edges. Conversely, if semi-natural habitat acts as a reservoir for FAW from which it colonises crops, this may be a more important determinant of FAW abundance, and FAW damage could therefore be higher closer to field edges.

FAW damage in relation to distance from field edge
To assess how herbivory from FAW varied with distance from field margins in a rural smallholder agricultural setting, on 5 February 2019 we surveyed two fields of unharvested maize bordered by semi-natural habitat (grass, shrubs, and small trees) in Abutia Amegame (Volta Region of Ghana; 6.46°N, 0.32°E), referred to throughout as village maize (VM). We sampled along a 7 m transect running from the mid-point of a field edge into the centre of the maize crop ( fig. 1a). Maize plants within a 2-m wide zone along this transect were surveyed for the total number of leaves per plant and number of leaves showing herbivory. The proportion of leaves with herbivore damage correlates with FAW abundance (Hruska and Gould, 1997;Midega et al., 2017). FAW infestation had occurred earlier in the growing season, so no FAW were present in the fields when surveyed, but the damage recorded was characteristic of FAW damage; i.e. windowed whorls with larval frass and skeletonised leaves (Goergen et al., 2016). We were unable to sample additional fields as intended due to the majority of maize fields having already been harvested.
To observe the infestation process of FAW on young maize plants in a more controlled experimental setting, on 5 February 2019 we established a 42 m × 7 m irrigated maize plot at the University of Ghana experimental farm (5.66°N, 0.19°W), referred to here as University maize 1 (UM1). The plot was bordered by bare cultivated earth on the long edges and semi-natural vegetation (predominantly grass) or a dirt access track on the short edges ( fig. 1b). We ran a 50 m transect starting 8 m into seminatural vegetation at one margin of the crop, through the whole experimental crop plot, and ending at the opposite margin bordering an access track. During 12-22 February 2019, we daily monitored 6-8 maize plants every 5 m along the transect in the two planting rows on either side of the transect. We recorded the total number of leaves and the number of leaves showing herbivore damage for the same plants each day. The number of FAW larvae per plant was not observed directly; this would have required destructive sampling, since FAW larvae typically retreat into leaf whorls (Day et al., 2017;Early et al., 2018).

Natural enemy activity in relation to distance from field edge
To link the development of FAW-mediated crop damage in young maize to natural enemy activity, richness and diversity, between 12 and 22 February 2019 we placed pitfall traps every 5 m along the 50 m transect in the University maize plot described above (UM1, fig. 1b). Traps were emptied every 24 h and invertebrate morphospecies abundances recorded. The four pitfall traps closest to the access track were excluded from analysis, due to inability to determine whether invertebrates in these samples had entered the crop from the edge of interest (bordering seminatural habitat) or had crossed the access track to reach the maize. We also assessed parasitism rates of FAW larvae from a three plant by three plant quadrat (totalling nine plants) every 5 m along the transect within the crop. Seventeen days after planting the maize, each plant was searched systematically for FAW larvae, and all larvae were collected (up to a maximum of 20 larvae per quadrat). Each larva was reared individually on maize leaves in transparent plastic cups covered with nylon mesh in the Department of Crop Sciences, University of Ghana, until either an adult moth or a parasitoid emerged, or the larva or pupa died; monitoring for parasitoid emergence continued for 1 week after death. Further collection of larvae took place at 14-day intervals, on 8 and 22 March 2019 (Supplementary  Table 4).
We complemented our study of natural enemy activity in the young maize crop by surveying previously established maize at the tasselling stage on the experimental farm (University maize 2, UM2). Two un-irrigated maize plots 28 m × 45 m had been planted on 9 November 2018. One of the 28 m margins of both plots was bordered by uncultivated land, predominantly exposed ground with patchy shrubs and grass clumps. One plot had previously been treated with a pesticide targeting FAW, but as pesticide treatment was not ongoing and is not of interest in our study, we sampled both plots to increase replication and included plot ID as a fixed effect in data analysis. We set a 35 m transect perpendicular to the edge bordering uncultivated habitat in both plots, starting 20 m into the semi-natural vegetation and continuing 15 m into the maize plot ( fig. 1c). On 11 February 2019, we placed pitfall traps (6 cm in diameter) partially filled with water and detergent, every 5 m along this transect. We emptied traps after 24 h and recorded invertebrate morphospecies. As a further indicator of natural enemy activity, ten plasticine caterpillars were placed every 5 m on the transect on 12 February 2019. Model caterpillars made from modelling clay or plasticine are widely used to assess potential predation pressure on invertebrate herbivores from arthropods, birds and mammals (Howe et al., 2009;Low et al., 2014). Dummy caterpillars (30 mm × 3 mm) were made from brown Newplast (Newclay Products Ltd), to mimic later instars of FAW (EPPO, 2015;Jeger et al., 2017). Caterpillars were attached to maize/semi-natural vegetation using UHU super glue. After 48 h, we used a hand lens to score the presence or absence of attack and the identity of the potential predator (categorised as arthropod, mammal, or bird) (Low et al., 2014). Mammal and bird attacks were rare (two and one caterpillars, respectively), so analyses were restricted to overall attack rates. Missing caterpillars were excluded from the analysis (as in Sam et al., 2015;Mansion-Vaquié et al., 2017).

Statistical analysis
R version 3.5.2 was used for statistical analysis and to plot graphs. Details of the variables and response distribution included in the generalised linear models (GLM, fitted using the glm function), along with details of dispersion tests and adjusted R 2 used to assess model fit are given in Supplementary Table 1 and Supplementary Text 1.
Araneae (spider) and Hymenoptera (mostly Formicidae, i.e. ant) data from pitfall traps were analysed as likely natural enemies of FAW. Three metrics were used to summarise Araneae and Hymenoptera activity and diversity: abundance (total number of individuals), species richness (number of species), and inverse Simpson's diversity index.

FAW crop damage
The proportion of leaves showing herbivore damage in the VM fields increased significantly with distance from semi-natural habitat by 4.2 and 3.9% per metre in the two fields (P < 0.05, Table 1, fig. 2a). However, relatively little of the variation in herbivory was explained by distance (R 2 D, Dγ = 0.0761, Supplementary Table 2).
In the UM1 plot, the relationship between level of herbivory and distance from semi-natural habitat changed with time since planting (i.e. a significant time × distance interaction; P < 0.001, Supplementary Table 2). Our hypothesis was primarily concerned with the effect of distance on herbivory, but since leaf damage accumulates as FAW infestation progresses, we included time as a fixed effect to account for this and to determine any interaction effect with distance. On the first day of sampling (7 days after planting), the proportion of leaves showing herbivore damage increased 0.078% over 10 m (15-25 m) from the semi-natural habitat, whereas 17 days after planting the proportion of leaves damaged decreased 0.59% over this same distance (Table 1, fig. 2b). The proportion of leaves with herbivory increased with time since planting from 12.0% 7 days after planting to 89.6% on day 17 (Table 1), but the increase in damage was less pronounced further from the field margin bordering semi-natural habitat ( fig. 2b). The binomial GLM had a very good fit to the data (R 2 D, γ = 0.476, Supplementary Table 2), equivalent to a conventional R 2 value of over 0.9 (Louviere et al., 2000).
Natural enemy activityabundance, richness, and diversity In UM1, a total of 2138 invertebrates were collected in pitfall traps over 11 days of sampling. Araneae accounted for 169 individuals assigned to 33 morphospecies, while a further 1670 were Hymenoptera, predominantly Formicidae (ants) and attributed to 64 morphospecies. The samples were dominated by three Araneae morphospecies (each with over 20 specimens across all days of sampling) and seven Hymenoptera morphospecies (combined abundance greater than 100). Araneae activity was affected by an interaction between distance from semi-natural habitat and time since planting (P < 0.05 for total abundance, P < 0.1 for richness and diversity, Table 2). As the maize grew, Araneae activity changed from moderately increasing with distance from semi-natural habitat to decreasing with distance ( fig. 3a-c). The total abundance data contained an outlier at 20 m on day 7 (abundance of 18) and re-analysis without this data point removed the statistical significance of this result, but not the direction of the effect (Supplementary Table 3). Araneae diversity was also significantly higher in semi-natural habitat than in the crop (P < 0.05, Table 2). Hymenoptera diversity was affected by an interaction between distance and time since planting (P < 0.05, Table 2), declining with increasing distance from semi-natural habitat into crop initially but, over time, changing to an increase with distance from crop edge ( fig. 3f). All GLMs used to analyse the Araneae and Hymenoptera metrics from the UM1 pitfall trap data explained little variation in the data, indicated by low R 2 values (Supplementary Table 3).
From the 260 FAW larvae collected at three 14-day intervals from UM1, 53 emerged as moths and eight as parasitoids, giving an average parasitism rate of 13.1% (Table 3) which is typical of parasitism of FAW in Ghana (Agboyi et al., 2020;Koffi et al., 2020) and East Africa (Sisay et al., 2018(Sisay et al., , 2019. Most other larvae died as larvae or pupae (Supplementary Table 5). Future work Hymenoptera species richness and diversity significantly increased with distance along the transect (P < 0.05, Table 4, fig. 4e, f), but did not differ significantly between habitat types. Statistical models used for analysis of Hymenoptera metrics fitted the data well, indicated by high R 2 values (Supplementary Table 3); total abundance and species richness models have adjusted pseudo-R 2 values in the range 0.2-0.4, equivalent to conventional R 2 values of 0.7-0.9 or above (Louviere et al., 2000). In contrast, the models used to analyse Araneae metrics explained little variation in the data (Supplementary Table 3).
Attack rate of dummy caterpillars placed in UM2 increased significantly with distance from crop edge in both the crop and semi-natural habitat (P < 0.05, Table 4, fig. 4g). Attack rate was significantly different between plots (P < 0.01). The model fitted very well, explaining a substantial amount of the variation in the data (R 2 D, Dγ of 0.446, Supplementary Table 3).

Discussion
We found that apparent FAW herbivory on maize (mature at the point of observation) increased with distance from semi-natural habitat in the realistic conditions of smallholder fields (VM). Our daily monitoring of herbivory on young maize plants at the University of Ghana research farm (UM1) indicated that FAW damage changed from increasing to decreasing with distance from semi-natural habitat as the maize grew. Araneae activity, richness and diversity decreased with distance from seminatural habitat over time in this crop, whilst Hymenoptera diversity increased. Further data from mature maize plots at the research farm (UM2) also found increased Hymenoptera richness and diversity, and dummy caterpillar attack rate, with distance from uncultivated ground at the crop edge. The maize plots on the university farm were relatively small and the semi-natural habitat bordering these plots is not consistently representative of semi-natural habitat surrounding fields in rural villages, which can include complex matrices of grass, shrubs and small trees. We did not record the maize varieties grown or plant species composition of the semi-natural habitat bordering the plots in this study, which would have allowed us to further qualify our findings and evaluate the prevalence of alternate hosts for FAW surrounding crop fields. Our study suffers from low replication, the magnitude of the effects we identified was very small and the differences between the VM and UM sites make comparison of results difficult. However, our results are useful as preliminary indicators of trends in a data-poor region. There are few published studies from West Africa that consider the relationship between seminatural habitat and natural enemy activity. A limited number of studies from the Americas consider the influence of extra-field characteristics (Wyckhuys and O'Neil, 2007), weeds within crops (Altieri and Whitcomb, 1980), and distance from semi-natural habitat (forest) (Sousa et al., 2011) on natural enemy and FAW abundance. Successful biological control of crop pests can be highly context-specific (Tscharntke et al., 2005;Bianchi et al., 2006), so regional studies of FAW in sub-Saharan Africa are necessary to fill this important knowledge gap. Future research would benefit from ongoing sampling of a greater number of fields throughout the maize growing season, with some measure of surrounding habitat and landscape complexity included in analyses, as this has been shown to be significant for other crop pests in sub-Saharan Africa (Kebede et al., 2018a(Kebede et al., , 2018b. While there are no published data from Africa, FAW abundance has been found to increase with distance from semi-natural habitat in the Americas (Sousa et al., 2011), consistent with our VM results. This trend could be due to edge effects resulting in decreasing FAW abundance closer to a non-preferred habitat (Ries et al., 2004) or natural enemy pressure being greater near the field edge due to spillover of enemies from adjacent non-crop habitat (Bianchi et al., 2006). In contrast, the pattern of herbivore damage in UM1 is consistent with crop pests infesting the field from the edge, re-colonising following post-harvest clearance (Tscharntke et al., 2005). The distance-dependent difference in herbivory emerged from around day 10 ( fig. 2b), corresponding to when FAW infestation is likely to have commenced; FAW frassa characteristic sign of FAW infestation (Hruska and Gould, 1997;Goergen et al., 2016;Jeger et al., 2017) first appeared on leaves 11 days after planting. However, it is unlikely that the trend in herbivory we observed is due to a gradient in FAW infestation from the crop edge. Although some pest species may benefit from specific components of semi-natural vegetation (e.g. Baggen et al., 1998), the primary source of most pest species are crop fields (Bianchi et al., 2006) and FAW adult females disperse long distances and oviposit directly onto maize throughout the field (Goergen et al., 2016). Crop yield measurements would provide a more robust and economically relevant measure of the impact of FAW. The yield loss from leaf damage depends on the maize growth stage; early infestation causes the largest yield losses, but yield loss plateaus with increasing pest pressure in later growth stages (Evans and Stansly, 1990;Overton et al., 2021). Other limitations of using leaf damage as a proxy for FAW abundance, particularly in the mature maize of the VM fields, include (i) inability to determine the maize growth stage when FAW attack occurred, (ii) difficulty distinguishing FAW leaf damage from other Lepidopteran crop pests, and (iii) leaf damage remaining in instances where FAW are removed by natural enemies. Despite its weaknesses, we feel leaf damage provided an adequate proxy to initially test our hypotheses and would seek to measure crop yields in future work. Table 2. Values of natural enemy metrics from UM1 pitfall trap data, evaluated as the difference between the 7 and 12 m sampling points from the crop edge for the first and last day of sampling (7 and 17 days after maize planting) Natural enemy metric n a Difference in natural enemy metric between 7-and 12 m from crop edge Significance of variable on natural enemy metric The decrease in Araneae abundance with increased distance from edge in UM1, and similar findings in early samples of Hymenoptera, are consistent with theoretical and empirical evidence that natural enemies colonise crops from surrounding semi-natural habitat in fragmented landscapes (Kruess and Tscharntke, 2000;Tscharntke et al., 2005;Bianchi et al., 2006;Sousa et al., 2011). Semi-natural habitats can enable the persistence of natural enemies during inter-crop periods when pests are not present by providing: (i) a source of alternative plant hosts and prey and nectar sources (important for parasitoids), (ii) a favourable microclimate, (iii) refuge, and (iv) sites for hibernation (Tscharntke et al., 2005;Bianchi et al., 2006;Harrison et al., 2019). The subsequent reversal in trend for Hymenoptera in UM1consistent with the trends in Fig. 3. Natural enemy activity (abundance, richness and diversity) data from pitfall traps in UM1; 8 m into semi-natural habitat (from crop edge) to 22 m into irrigated maize plot. Lines display model-fitted values, points show the raw data, and colours indicate day since maize was planted (11 days surveyed in total). (a) Araneae abundance (total number of individuals), (b) Araneae species richness (number of species), (c) Araneae diversity (inverse Simpsons Diversity Index, log scale), (d) Hymenoptera abundance (total number of individuals, (e) Hymenoptera species richness (number of species), (f) Hymenoptera diversity (inverse Simpsons Diversity Index). As the maize grew, Araneae activity tended to decrease with distance into the maize plot, whereas Hymenoptera activity increased.
Hymenoptera and dummy caterpillar attack rate in UM2is expected if natural enemies have moved from the semi-natural habitat and stayed in the crop as the maize grew, perhaps facilitated by the increased suitability of the crop as habitat and/or availability of crop pests as food sources. If this is the case, it is possible that Araneae are slower to move into the crop following the arrival of crop pests, or require a more developed crop to provide a suitable matrix, potentially explaining why we detected this trend reversal for Hymenoptera but not Araneae in UM1. Our results therefore highlight the influence of time since planting on the distribution of potential natural enemies with distance from semi-natural habitat. Early FAW infestation of maize results in the largest yield losses (Evans and Stansly, 1990), and our study found evidence of FAW infestation 11 days after planting. As such, natural enemies from semi-natural habitat surrounding crops may not effectively minimise yield loss if they do not disperse into the crop rapidly enough.
FAW infestations in sub-Saharan Africa are causing substantial damage to maize crops (Sisay et al., 2019) with severe yield consequences (Njuguna et al., 2021) in a context where food security is already of concern (FAO et al., 2019). Control of FAW to date is focusing almost exclusively on the use of synthetic insecticides (Harrison et al., 2019;Agboyi et al., 2020;Tambo et al., 2020) despite recognition of the negative impacts of pesticide application on human health and the environment (Kansiime et al., 2019;Tambo et al., 2020;Haggblade et al., 2021). Enhancing landscape complexity by increasing the proximity of crops to semi-natural habitat has been suggested as a means to increase natural enemy predation of crop pests, both generally (Bianchi et al., 2006) and specifically for FAW (Harrison et al., 2019), thus reducing the need for synthetic insecticides as part of an integrated pest management approach. Our results support further investigation of changing maize field configuration to increase proximity of crops to semi-natural habitat (e.g. through dividing larger fields with semi-natural vegetation strips), to reduce FAW damage and increase natural enemy abundance and diversity within crops. Future work should assess the impact of such a strategy on maize yield, to ensure reductions in FAW damage translate into improved or safeguarded yields that more than compensate for potential loss of cropping area from habitat enhancement.

Conclusion
This study provides the first quantification of changes in FAW crop damage and activity of its potential natural enemies with distance from semi-natural habitat bordering maize crops in West Africa. We observed increased herbivore damage with distance from semi-natural habitat in a small sample of  Values given are back-transformed from models fitted. Significance of variables on natural enemy metric are also given (-not significant, . P < 0.1, * P < 0.05, ** P < 0.01). For the full summary of model estimates, standard errors and R 2 values, see Supplementary Table 3. a Eight intervals on the transect were sampled over across two plots, resulting in n = 16. Samples with zero individuals were removed from analyses of inverse Simpsons index.
smallholder maize fields and provide initial evidence that abundance of some natural enemies declines with distance from semi-natural habitat in young maize around the time of FAW infestationa particularly critical stage for reducing pestmediated yield loss. Further work is required to quantify the extent to which natural enemy pressure reduces FAW abundance, and whether this occurs rapidly enough to prevent significant yield loss due to herbivore damage, before the effectiveness of such management practices as a means of improving food security can be assessed.
Supplementary material. The supplementary material for this article can be found at https://doi.org/10.1017/S0007485321000894