The Koga Irrigation Scheme (KIS) is located in the Upper Blue Nile Basin at 11° 10’and 11°25’ North and 37°2’ and 37°17’ East in Mecha district, Amhara National Regional State, Ethiopia. The KIS potentially irrigates a 7000-ha command area using rainwater harvested in a dam from a 22,000-ha catchment in the upper Blue Nile Basin, which drains into the Koga River (Asres, 2016). It is Located approximately 41 km south of Bahir Dar City along the route to Addis Ababa, with an average altitude of 1953 m above sea level. The mean maximum and minimum temperatures are 26.8°C and 9.7°C respectively. Mean annual rainfall, recorded at the station of Merawi (the main town of Mecha district), was 1480 mm, with 90% occurring from May to October. The dominant soil type in the survey area was Nitisol with strong acidity (pH 5.04 -5.25) and it has medium to high organic matter (2.34% - 4.44%) content. The available phosphorus is moderately good (53–55 ppm) while it has medium total nitrogen content (0.2-0.23%) (Asmamaw et al., 2021).

Pest surveys were conducted in maize fields at KIS during the main cropping season in July 2019 and in the irrigation season December, 2019. A total of 123 randomly selected maize fields were surveyed across KIS, including 62 fields during the main season and 61 during the irrigation season (Figure 1). The study area encompassed eight irrigation blocks (command areas): Tagel, Chihona, Ambomesk, Tekledib, Bered, Andinet, Amarit, and Lasi. Data were recorded on agronomic and field characteristics, including previous crop, maize growth stage, and weed management as well as pest parameters, such as incidence, and severity expressed as leaf damage percentage (LDP), LDP by damage type (scratching, pinhole, and rugged), and larval instar stages observed on sampled plants. ‘Previous crop’ refers to the crop planted in the specific field during the preceding season. ‘Weed management’ was assessed based on critical weeding periods (Gantoli et al., 2013), and categorized as good, medium, or poor. Fields classified as “good” were weeded twice, once at the 4–6 leaf stage and again at the 8–10 leaf stage, and had an average of <15 weeds m^-². Fields weeded only once during either period were classified as “medium” and had an average of 65 weeds m^-². Fields with no weeding until the survey date were classified as “poor” and had >120 weeds m^-².

The FAW ‘Incidence’ was calculated as the percentage of plants showing FAW infestation relative to the total number of sample plants per field (n=50). FAW severity was recorded based on ‘Leaf damage percentage’ (LDP) referring to the mean percentage of leaves damaged by FAW infestation as indicator of both host preference and host suitability. LDP by damage type was recorded by classifying leaf damage percentage into three FAW-specific categories: scratching, pinhole, and rugged (Supplementary File 1: Supplementary Figure S1).

In each field (< 2 ha), a total of 50 plants were randomly selected and assessed for FAW incidence and severity. For fields in vegetative stages up to late whorl, plants were sampled using a ‘W’ scouting pattern; for fields in reproductive stages, a ‘ladder’ pattern was used. All 50 sampled plants were individually examined for FAW incidence and severity using the LDP, and larval instars were recorded when live larvae were present.

The host preference and damage progression of FAW relative to its primary host, maize (Zea mays L.), were evaluated on four cereal hosts commonly grown in the region: finger millet (Eleusine coracana L. Gaertn.), Wheat (Triticum aestivum L.), barley (Hordeum vulgare L.), and tef (Eragrostis tef [Zucc.] Trotter). The cage experiment was conducted under controlled conditions at the Amhara Agricultural Research Institute, Bahir Dar, Ethiopia following the institutional biosafety guidelines. Each experimental unit consisted of a plastic pot (25 cm top diameter × 15 cm bottom diameter × 20 cm height) filled with a standard soil mixture and planted with one crop species. Immediately after germination, pots were transferred to insect-proof cages (0.75 × 0.75 × 1 m, L × W × H) to prevent external infestation. Host preference tests were conducted using a completely randomized design with three replications per treatment. For each replication, the positions of the cages within the experimental area and the assignment of pots to cages were randomized. In the choice test, four pots—three containing one of the test cereals (finger millet, barley, wheat, or tef) and one containing maize as the preferred reference host were placed together in each cage. In the no-choice test, four pots each containing a single test cereal were arranged within a cage in the absence of maize. All cages were maintained in an open, unshaded area to ensure natural temperature, humidity, and photoperiod conditions.

S. frugiperda larvae collected from naturally infested maize fields were reared under laboratory conditions until adult emergence. Newly emerged adults of uniform age were selected for release in the experimental cages. In each cage, two female and one male adult moths were introduced and the cages remained sealed throughout the experimental period to prevent escape or external contamination. Egg masses laid by the released adults were monitored daily until hatching. After hatching, larval feeding damage on each host plant was assessed at 1–3-day intervals for 30 days after hatching (DAH) to evaluate both the damage progress and host feeding preferences (Supplementary File 1: Supplementary Figure S2).

Map of the study area was constructed using ArcMap version 10.7.1. The shape files were constructed using QGIS Desktop version 3.38.0 and the base maps were from the ArcMap 10.7.0 inbuilt sources. Google Earth Pro software was used to display sampling points on the retrieved areal image of Mecha district and KIS command areas from Airbus 2024. Descriptive statistical analysis was carried out using SPSS version 25. Parameter estimates of the logistic regression were analyzed with the PROC GENMODE model in SAS version 9.0 to evaluate multiple explanatory variables and comparison across different variable classes. Prior to data analysis, normality test was made using Shapiro-Wilk test and the homogeneity of variances was determined using Levene’s test. Then, statistical tests on differences of FAW incidence and LDP in the main and irrigation season were tested using Mann-Whitney U test and independent Student’s t-test, respectively, using R software version 4.4.3. Damage percentage data (Supplementary Files 2, 3) were analyzed using a linear mixed-effects model (LMM) implemented in the lme4 version 1.1-37 (Bates et al., 2015) and lmerTest version 3.1-3 (Kuznetsova et al., 2017) R packages. Model parameters were estimated using restricted maximum likelihood (REML), and the significance of fixed effects was assessed using Type III ANOVA with Kenward–Roger degrees of freedom correction for maize-cereal co-cultivation, while Kruskal-Wallis test followed by Dunns’s test was performed for cereal co-cultivation data. Estimated marginal means (EMMs) for crop damage × DAH combinations were computed using the emmeans package version 2.2.0 (Searle et al., 1980), and pairwise comparisons among host crops at each observation interval were adjusted using the Tukey method to control for multiple testing. Compact letter displays were generated to denote statistically distinct groups. Temporal patterns of damage progression were visualized by plotting EMMs across days for each crop using ggplot2 package version 3.5.1 (Wickham et al., 2016). The SEM was used to assess the direct and covarying effects of experimental factors on fall armyworm incidence and leaf damage percent. Two SEM models were tested: the first evaluated the influence of cropping season, maize growth stage, insect developmental stage, and leaf damage type on incidence and LDP, with both outcomes allowed to covary; the second assessed the effects of previous crops on incidence and LDP. All numeric data were standardized before the analysis, and the SEM correlation matrix was processed using R software version 4.4.3 using lavaan package version 0.6-20 (Rosseel, 2012) and semPlot package version 1.1.6 (Epskamp, 2015).

The cumulative prevalence of FAW infestation in both the main and irrigation seasons in KIS maize fields reached 99.2%. Maize was cultivated continuously across seasons in 8.9% of surveyed fields, while 86% were rotated with eight different crop types, dominated by wheat and potato; 4.8% of fields were fallow in the preceding season. Most maize fields were at early to late whorl stages (V3–VT), with 65% infested by 3rd–5th instar larvae. However, 30.9% of fields were dominated by 6th instar larvae. FAW incidence was considerably higher during the irrigation season, ranging from 26% to 96%, compared with 0% to 60% in the main season. This seasonal difference was reflected in the level of crop injury with average leaf damage reached 60% in the irrigation season, whereas it was only 33.6% during the main season (Table 1).

The Mann-Whitney U test confirmed that FAW incidence in the irrigation season (median = 66%) was significantly higher (p < 0.001) than in the main season (median = 24%) (Figure 2A).

An independent Student’s t-test confirmed significantly higher FAW leaf damage percentage (LDP) during the irrigation season (mean = 59.98%, n = 61) compared with the main season (mean = 33.55%, n = 62; P < 0.001). (Figure 2B). Damage type also differed between seasons: rugged damage dominated in the irrigation season (39%), while pinhole damage was more common in the main season (12.3%) (Table 1).

Logistic regression analysis (Table 2) showed that maize fields during the irrigation season had a significantly higher likelihood of fall armyworm (FAW) incidence compared to those in the main cropping season (OR = 4.30, p < 0.001). Previous crop type also emerged as a significant predictor of FAW incidence in the subsequent season (p < 0.01), with distinct risk profiles across preceding crop types (Table 2). Fields previously planted with snow pea (Pisum sativum), tomato (Solanum lycopersicum), and finger millet (Eleusine coracana) had a significantly higher risk of FAW incidence (p < 0.001) than in onion-preceded fields, exceeding the risk of replanting maize (Zea mays) in consecutive seasons (OR = 5.58, 2.77, and 2.58, respectively). Maize-preceded fields were approximately 2.5 times more likely to experience FAW incidence in the following season than onion-preceded fields (OR = 2.48, p < 0.001).

Fields previously cropped with wheat (Triticum aestivum), barley (Hordeum vulgare), nug (Guizotia abyssinica), or potato (Solanum tuberosum) also showed significantly increased likelihood of FAW incidence (Table 2). Fallow plots, however, had no significant effect on FAW risk (OR = 1.30, p = 0.218), indicating limited pest carryover during non-host periods (Table 2). Weed abundance and maize growth stage were also identified as significant predictor variables for FAW incidence. Fields with low weed abundance showed a moderately higher likelihood of FAW infestation (OR = 1.32, p < 0.001), whereas fields with less frequent weeding tended to experience lower infestation rates. On the other hand, maize plants at the early whorl growth stage were significantly more susceptible to FAW attack, showing a slightly higher likelihood of infestation (OR = 1.13, p < 0.02; Table 2).

Analysis of parameter estimates for LDP (Table 3) revealed consistent seasonal trends, with irrigated maize fields showing an approximately threefold higher likelihood of severe leaf damage compared to fields cultivated during the main cropping season (OR = 2.9, p < 0.0001). Because the previous crop was a weak predictor of LDP overall, fields preceded by crops other than snow pea (Pisum sativum), nug (Guizotia abyssinica), or tomato (Solanum lycopersicum) did not show a significant effect on LDP. However, fields preceded by snow pea (P. sativum), nug (G. abyssinica), or tomato (S. lycopersicum) exhibited an approximately 40% lower likelihood of severe leaf damage compared with fields previously planted with onion (Allium cepa). Growth stage of maize, and stages of larval instars were also identified as significant predictors of LDP. Maize plants at the seedling stage exhibited a higher likelihood of severe leaf damage (OR = 0.06, p < 0.0001) compared with seedlings at the late whorl stage. Similarly, plants at the early whorl stage showed a significantly lower likelihood of LDP (OR = 0.09, p < 0.001) than those at the late whorl stage. Early larval stages (first and second instar) were associated with a greater likelihood of increased LDP than the sixth-instar larvae (OR = 1.4, p = 0.003).

The structural equation model (Figure 3) revealed that irrigation-season maize cropping was significantly and strongly associated with higher FAW incidence (path coefficient = 0.82, p < 0.001) and with increased LDP (LDP; path coefficient = 0.67, p < 0.001). Irrigation-induced damage was also significantly correlated with both ragged damage (path coefficient = 0.70, p < 0.001) and pinhole damage. On the other hand, the growth stage of the crop showed a negative association with FAW incidence (Figure 3). The SEM analysis on the relationships of previous crops with incidence and severity showed that Potato had the strongest positive effect on FAW incidence (β = 0.73, p < 0.001) followed by finger millet (β = 0.65, p < 0.001), and maize (β = 0.49, p < 0.001) and snow pea (β = 0.44, p < 0.001) (Figure 4). Potato and finger millet rotations were associated with substantially higher infestation than non-rotated maize. Despite strong association of previous crops with FAW incidence, LDP exhibited moderate patterns with a non-significant association with leaf damage percentage.

Host preference assays using larvae from two adult females under controlled cage conditions showed no host discrimination among maize, barley, and finger millet, all of which experienced complete destruction within 30 DAH (p > 0.05) (Supplementary File 1: Supplementary Table S1; Supplementary Figures S3A, B). In contrast, wheat exhibited moderate resistance and significantly lower damage than finger millet, barley, and maize (p < 0.001), while tef sustained minimal damage at 30 DAH (Supplementary File 1: Supplementary Figures S3C, D; Supplementary Table S1). Estimated marginal means confirmed these trends: finger millet and barley reached 100% damage, maize reached 98.3%, wheat showed 75% damage, and tef experienced the least damage (10%) when co-cultivated with maize throughout the 30-day observation period (Supplementary File 1: Supplementary Table S2).

Leaf damage progressed gradually over time, with the first noticeable symptoms appearing at 11 DAH. At this stage, damage across all cereals was minimal and not statistically different (p > 0.05; Figure 5; Supplementary File 1: Supplementary Table S3), consisting mainly of light leaf scratching, followed by a slight increase up to 15 DAH (Figure 4). Damage progression in finger millet, maize, and barley did not differ significantly until 16 DAH. In contrast, wheat exhibited significantly lower damage than finger millet, and no damage was recorded on tef at this time point. A sharp increase in damage occurred at 19–20 DAH, during which barley, finger millet, and maize showed statistically similar and substantially higher damage compared with wheat. Tef remained the least affected, with no measurable damage until 23 DAH (Figure 5). The damage progressively continued and at 30 DAH, complete plant damage was observed on barley, finger millet, and maize. Wheat showed significantly lower overall damage compared with these cereals, although its leaf tissue was fully consumed and surviving panicles contributed to a moderate damage estimate (Emmean = 75%; Figure 5; Supplementary File 1: Supplementary Table S2). Tef exhibited significantly lower damage than all other cereals when co-cultivated with maize. However, when cereals were co-cultivated in the absence of maize, complete damage occurred across all, including tef, finger millet, wheat, and barley, within 30 days after larval hatching (DAH) (Supplementary File 1: Supplementary Figures S4, S5).