Fall armyworm (Spodoptera frugiperda Smith) was first detected on the African continent in early 2016 (Goergen et al. 2016). It spread rapidly across sub-Saharan Africa, the Middle East and South Asia, reaching East Asia and Australia in 2018 (Nagoshi et al. 2020). S. frugiperda is a serious pest from the Americas, where it causes substantial damage to cereal crops especially maize and rice, and its arrival in the Old World threatens the food security and incomes of millions of smallholder farmers. In response, countries have released millions of dollars in emergency funding and donated chemical pesticides to farmers (Jepson et al. 2020). However, widespread use of toxic chemicals has serious consequences for human health and the environment (Lewis et al. 2016, Isgren and Andersson 2021, Tang et al. 2021). Poor farmers rarely use personal protective clothing, thereby exposing themselves to high levels of toxic chemicals when they spray fields, and pesticide containers are rarely disposed of properly and may even be reused as water-bottles or food containers. In addition, broad spectrum pesticides often have a much greater impact on natural enemies than on the target pest, which can lead to pest resurgence if farmers do not spray their fields multiple times a season (Hardin et al. 1995). This is particularly true for S. frugiperda, which feeds and hides in the maize whorl rendering many contact pesticides less effective. Furthermore, long-term use of chemical pesticides fosters a dependency, driven by their impacts on natural enemies and the development of pest resistance, leading to escalating pest management costs, which is something smallholder farmers can ill afford (Lewis et al. 1997). Hence, there is an urgent need to understand the impacts of S. frugiperda across the invaded range and to develop alternative pest management options that can be incorporated into a sustainable Integrated Pest Management (IPM) approach.

S. frugiperda is a highly polyphagus, highly mobile, and highly fecund pest. It has been recorded feeding on >350 species in its native range (Montezano et al. 2018) and hence can persist on a wide variety of host plants when its preferred hosts, primarily maize and other poaceous plant species, are not available. The adult moths can fly continuously for over 24 h and cover over 100 km through self-powered flight (Chen et al. 2022). It is therefore capable of quickly colonising crops over a wide area soon after the seedlings emerge. Last, female S. frugiperda moths lay egg masses of 50–350 eggs and can lay up to 1,500 eggs over their adult life of approximately 3 wk. Under optimal temperatures (∼25°C), S. frugiperda can complete its life-cycle within approximately 4 wk (Du Plessis et al. 2020) and hence, under suitable conditions, populations have the potential to increase rapidly. However, S. frugiperda is also attacked by a wide diversity of natural enemies, including parasitoids, predators, nematode parasites, entopathogenic fungi, and viruses. In its native range, around 200 species of parasitoid have been recorded attacking S. frugiperda (Molina-Ochoa et al. 2003) and already a substantial number of parasitoids have been recorded emerging from S. frugiperda in the invaded range (Sisay et al. 2018, 2019; Tendeng et al. 2019, Agboyi et al. 2020, Durocher-Granger et al. 2020). While in the Americas, S. frugiperda management has mostly focused on chemical pesticides and Bt corn, reports suggest that natural enemies can effectively control the pest under certain circumstances. For example, high proportions of S. frugiperda larvae may be attacked by parasitoids in unsprayed fields (Meagher et al. 2016) and significant mortality from fungal infections has been recorded (Guo et al. 2020). Heavy rainfall is also an important mortality factor for S. frugiperda (Varella et al. 2015). Young larvae are washed off leaves and older larvae drown in the whorl. Importantly, in Central America, smallholder farmers who manage biodiverse plots known as milpas, do not consider S. frugiperda a serious problem although it is present in most fields (Wyckhuys and O'Neil 2007). Ultimately, whether the expected yield losses surpass economic action thresholds will depend on the balance between pest pressure and natural mortality factors. Unfortunately, monitoring efforts and guidelines for action thresholds often only focus on the former (e.g., McGrath et al. 2018) and ignore the important role of natural mortality.

Since its arrival in sub-Saharan Africa, there have been a number of reports concerning the impacts of S. frugiperda. Most have suggested that these impacts are substantial (Kansiime et al. 2019, De Groote et al. 2020, Kassie et al. 2020, Abro et al. 2021, Tambo et al. 2021, Kwasi et al. 2022, Makale et al. 2022). For example, based on household surveys, Tambo et al. (2021) reported that severe infestations caused a 44% reduction in household income and a 17% increased likelihood of experiencing hunger in Zimbabwe. Similarly, in Kenya De Groote et al. (2020) used group discussions to assess the impact of S. frugiperda, and reported yield losses of 30% nationwide and over 50% in some areas. Using a similar approach in Ethiopia, Abro et al. (2021) concluded that S. frugiperda caused a 36% reduction in maize production nationwide. However, these estimates are all based on farmer perceptions, without any independent verification of impact estimates from field data. Even where field assessments were made many only extend to assessing infestation and leaf damage (Sisay et al. 2019), which may not be reliable predictors of yield loss (e.g., Baudron et al. 2019). Maize has a remarkable capacity to recover from leaf damage (Hruska 2019). Even under controlled screen-house conditions, it is often difficult to predict yield loss from S. frugiperda infestation or leaf damage estimates. In farmers' fields the situation is made more difficult as a result of the variation in yields generated by differences in soil quality, fertilisation rates, weed management, and the effects of other pests and diseases. Although on-station trials indicate yield losses can be severe when pest pressure is high (Van den Berg et al. 2021), there are few published estimates of yield loss due to S. frugiperda in sub-Saharan Africa based on field measurements. Moreover, those that exist tend to suggest a lower impact than the above mentioned studies. For example, Baudron et al. (2019) surveyed 791 smallholder fields in Zimbabwe and estimated a yield loss of 11.6% from S. frugiperda. Quality information concerning the impact of S. frugiperda is critical for designing and implementing IPM strategies, as well as for allocating limited national agricultural budgets to best effect. To address this critical knowledge gap we implemented a large-scale on-farm trial across a wide climatic gradient in Malawi and Zambia. Here, we report on the results from our conventional tillage, monoculture maize, no pesticide treatment, which served to monitor the background S. frugiperda impact in the absence of control measures.

Materials and Methods

Data Collection

Scouting data were collected twice each season approximately 3 wk apart. Because of the challenges of covering so many locations, the timing of scouting varied, and hence we categorised it into early and late scouting periods (3–6 wk and 6–9 wk after emergence, respectively). All data were collected from the plot core area and we used a W-format to determine data collection points. At each point on the W, we collected data from 10 plants (i.e., 50 plants total per plot), including plant height (in cm), leaf chlorophyll content (SPAD units, Konica Minolta SPAD502-plus), S. frugiperda infestation (Yes, No), leaf damage score using the Davis scale (Davis et al. 1992, Toepfer et al. 2021) (score 1–9), and presence of other pests and disease, such as stemborers or Maize streak virus (Yes, No). We also recorded some plot level information, such as the condition of the plot with respect to weeding. The data on plant height and leaf chlorophyll content were used to provide an objective assessment of plant health status.

At harvest, we collected data from 2 m subsamples of each row. The 2 m subsamples were off-set so that we collected a representative sample diagonally across the plot core area. For each 2 m strip, we recorded the number of plants and plant height (in cm), the number, length (in cm), and mass of cobs (in g) on each plant, the ear damage score for each cob (CIMMYT scale, score 1–9) (Prasanna et al. 2018), the raw grain weight per plot (in kg) and the adjusted grain weight per plot (in kg). To calculate the adjusted grain weight, we took a sample of grain from each plot and dried it in an oven to constant mass to calculate the moisture content and then adjusted the weight to the standard 12.5% moisture content.

Table 1.

Description of sites including mean annual rainfall, the maize variety and fertiliser inputs used, and the numbers of farmers participating in each of the high and low tree cover landscapes at each site in the 2019/2020 and 2020/2021 maize seasons

Results

Infestation Rates

S. frugiperda infestation rates varied substantially among sites, landscapes, and seasons (Fig. 1,  Supp Table 1 [online only] (toac113_suppl_supplementary_material.zip)). In the 2019/2020 season, the proportion of plants infested with S. fruigiperda in a plot varied from 0.00 to 1.00, while the median plot-level infestation varied from 0.00 to 0.57 across landscapes. The corresponding figures for the 2020/2021 season were 0.00–0.69 and 0.00–0.32. Infestation rates were lower in high tree cover landscapes, in the late season scouting, and in the 2020/2021 season (Fig. 1,  Supp Table 1 [online only] (toac113_suppl_supplementary_material.zip)). In addition, there were significant interactions between rainfall and year, rainfall and scouting time, landscape tree cover and year, and landscape tree cover and scouting time. Thus, S. frugiperda infestation was always low at Kawambwa, which was the highest rainfall site, but the pattern across other sites was inconsistent. The effects of scouting time and landscape tree cover were reduced in the second season, apparently reflecting the overall reduction in infestation rates.

Fig. 1.

Proportion of infested plants per plot (S. frugiperda (FAW) incidence) across landscapes with high and low tree cover. Scouting was conducted at 3–6 wk (early) and 6–9 wk (late) in the 2019–2020 (2020) and 2020–2021 (2021) seasons. Sites are arranged from left to right in order of increasing mean annual rainfall. Missing plots/bars indicate missing data.

Discussion

Our dataset is derived from a large-scale trial, covering a wide range of rain-fed maize growing conditions. Our trials were established across six sites in Malawi and Zambia, with two landscapes per site and 12–15 farms per landscape, and repeated over two years. Plots were established using standard protocols and we used appropriate seed varieties and fertiliser applications for the local conditions. In short, this was a well replicated, large-scale trial that is well placed to assess the impact of S. frugiperda on smallholders' fields in Malawi and Zambia.

We found that S. frugiperda infestation levels varied from zero to 1.00 at the plot scale, while median of plot infestation rates varied from zero to 0.57 across landscapes. At the high end of the scale, our results fall within the range of values reported from around the region. However, in our trial a large number of plots had very low infestation (Fig. 1). We also found low levels of leaf damage in most plots (Fig. 2). Maize plants have a high capacity to recover from leaf damage (Hruska 2019) and hence leaf damage scores of less than Score 4 (<5 on the original Davis scale) are unlikely to have any impact on yield. A more informative metric may be the proportion of severely damaged plants (Table 2). However, 84% of our plots did not have any severely damaged plants and only 3% of plots had ≥10% severely damaged plants. This suggests that a large proportion of S. frugiperda larvae are dying before they reach the fifth and sixth instars, which cause most of the damage (at low levels of infestation larvae may also move among plants). Unsurprisingly, therefore, infestation rate was not a reliable predictor of severe leaf damage (r² = 0.06).

Fig. 2.

Plot mean leaf damage score per plot across landscapes with high and low tree cover. Damage score was assessed on the Davis 1–9 scale and rescale from 0 to 8, so that 0 = no damage. Scouting was conducted at 3–6 wk (early) and 6–9 wk (late) in the 2019–2020 (2020) and 2020–2021 (2021) seasons. Sites are arranged from left to right in order of increasing mean annual rainfall. Missing plots/bars indicate missing data.

Table 2.

Proportion of plots with different percentages of severely damaged plants (adjusted Davis score 4–8) across scouting periods and seasons

After controlling for variance among sites, we found that S. frugiperda infestation rates and leaf damage scores were lower in high tree cover landscapes, the late season scouting, and in the second year. In the second year the effect of scouting time was less pronounced, but this can be explained by the fact that the S. frugiperda infestation rates and damage scores were low and hence there was limited variance (Figs. 1 and 2). Rainfall, especially heavy rain, is known to be an important mortality factor for S. frugiperda (Varella et al. 2015). Young larvae may be washed off leaves and old larvae can drown in the whorl. Wet conditions may also promote higher rates of infection with entopathogenic fungi (Guo et al. 2020). Hence, finding that our wettest site had very low infestation and leaf damage aligns with expectations. In addition, we found that high tree cover landscapes had significantly lower rates of infestation and leaf damage. Landscapes with higher tree cover may be expected to deliver enhanced ecosystem services through harbouring more plant biodiversity (Wan et al. 2020). This may include insectivorous bats (Maas et al. 2016), which are important predators of the adult S. frugiperda moths, and arthropod natural enemies (Wan et al. 2020). Clarkson et al. (2022) also recently found that proximity to forest reduced S. frugiperda infestation in Zimbabwe. These results underline the importance of conserving areas of (semi-)natural vegetation in agricultural landscapes for pest control services. We recorded a decline in incidence of infestation and leaf damage severity from the early to the late season scouting. Variation in the proportion of larvae versus other stages of the life-cycle (i.e., pupae or adults) can drive temporal variation in infestation and damage, but as our sites support overlapping generations this is unlikely to have a substantial effect. Rather, the declining impact of S. frugiperda through the season is most likely caused by natural mortality.

We could not predict maize yield from infestation or leaf damage, even when we used data from both scouting assessments (Fig. 3). Maize has a tremendous capacity to recover from leaf damage (Hruska 2019) and it is often difficult to determine yield loss in carefully controlled screen-house trials, let alone on farmers' fields. Differences in soils among farmers' plots and among farmers in their management of the plots contributed to high background variance in yields, even though we standardised the planting protocols and fertiliser application rates. Other pests and diseases may also have contributed, but the rates of infestation with Maize streak virus (6.1%) and stemborers (0.2%) were low. On the one hand, this indicates that for smallholders far bigger yield gains can be achieved through improved soil management and crop husbandry, than through enhanced S. frugiperda control. On the other hand, this places us in a quandary when trying to develop action thresholds for IPM. Our results, and those from other studies (Baudron et al. 2019), indicate that infestation rates are not a reliable indicator of yield loss and, contrary to prevailing advice (McGrath et al. 2018), should not be used. In the 2019/2020 season, 40% of our plots exceeded the early season threshold for action (i.e., 20% infestation) and 17.6% exceeded the late season threshold for action (i.e., 40% infestation). In the 2020/2021 season the figures were 10% and 2.2%, respectively. However, very few if any of our plots suffered sufficient impact to justify spraying. A large, but variable, proportion of the S. frugiperda population occurring in a field will suffer from natural mortality, so the infestation rate imparts limited useful information. We suggest focusing on the proportion of severely damaged plants, because this serves not only as a measure of damage but also an index of the number of larvae that are surviving to the late stage instars. In addition, we found that infestation level and leaf damage declined from the first to the second scouting. Hence, action thresholds should probably be based on at least two scouting assessments. A useful metric could, e.g., be the trend in the proportion of severely damaged plants from the first to second assessment.

Fig. 3.

Yield per ha (kg) against (A) S. frugiperda incidence and (B) S. frugiperda leaf damage score for the 2019–2020 (2020) and 2020–2021 (2021) seasons. Circles and solid trend line indicate data from the early scouting period (3–6 weeks after emergence) and triangles and dashed trend line represent data from the late scouting period (6–9 weeks after emergence). Maize yield could not be predicted from S. frugiperda incidence or S. frugiperda leaf damage scores.

The levels of S. frugiperda infestation and leaf damage we recorded suggest a minimal impact on smallholders' fields in Malawi and Zambia. Against the background variance caused by differences in soils and plot management, we were unable to detect any significant effect of S. frugiperda infestation, or leaf damage on yield. These results are in stark contrast to many other reports from the region, including from Malawi and Zambia, that have tended to suggest very high levels of yield loss caused by S. frugiperda (Kansiime et al. 2019, De Groote et al. 2020, Kassie et al. 2020, Abro et al. 2021, Tambo et al. 2021, Kwasi et al. 2022, Makale et al. 2022). How can we reconcile these very different findings? Estimates based on farmer perceptions have been found to exaggerate S. frugiperda impacts (Baudron et al. 2019). In addition, the metric used in social survey based studies is often simply S. frugiperda affected versus non-S. frugiperda affected households (De Groote et al. 2020, Kassie et al. 2020, Tambo et al. 2021, Kwasi et al. 2022), which is clearly problematic. If S. frugiperda is present in an area, then at the scale of a smallholder property (∼1 ha) one would anticipate close to 100% infestation. Thus, reported results may reflect geographically driven variation in yields that have nothing to do with S. frugiperda. Alternatively, the farmers may be parsing the metric somewhat to separate heavily affected fields from less affected fields, but with unclear boundaries. However, if a field in a particular area is badly affected by S. frugiperda compared to neighbors, this may well be driven by poor management, such as late planting or poor weeding (Baudron et al. 2019). So it is then unclear whether the metric is measuring the impact of S. frugiperda or variation in management quality. In short, unless data on S. frugiperda infestation, damage, and yield loss are collected through field measurements, we cannot reliably assign differences in household income or food security to S. frugiperda.

More generally, although our study covered a wide range of agroecological conditions, it obviously did not encompass all possible situations. For example, we are aware of some ‘hot-spot’ areas in Zambia and Malawi. These appear to be where winter maize is grown using irrigation from permanent streams, such as in Chirundu district along the banks of the Zambezi, thereby maintaining a high population of S. frugiperda through the dry season. In other countries, where there are two cropping seasons or a single extended maize growing season, populations of S. frugiperda may be able to build up to high levels and escape their natural enemies. It is also possible that immediately following its arrival in sub-Saharan Africa the impacts of S. frugiperda were higher and that they have subsequently subsided as natural enemies have learnt to attack the pest. Unfortunately, we do not know because reliable data are lacking. Our results suggest the impacts of S. frugiperda may have been overestimated in many places across the continent.