Field trials were conducted for two consecutive years, starting in November 2020 in Indonesia and the March 2021 planting season in Indonesia and Thailand. Sixteen trials were conducted across the Java and Sumatra islands of Indonesia and Thailand’s key maize-producing provinces. The trials will be completed by the end of 2021. We collaborated with more than 500 growers to procure nearly 600 ha of land for these trials. The details of the locations are presented in Table 1.

These large-scale studies evaluated the performance of slow-release low-density polyethylene (LDPE) dispensers filled with 2.5 grams of blended FAW sex-pheromone active ingredients, (Z)-9-tetradecen-1-yl acetate (Z11- 16Ac), and (Z)-11-hexadecen-1-yl acetate (Z11- 16Ac) in a ratio of 87:13. Dispensers were installed on thin bamboo sticks with a cable tie 0.5 m above the ground. Pheromone bucket traps were used to trap the male moths and to evaluate MD via trap reduction. Commercially available FAW lures with similar blend ratios and bucket traps were purchased from Pherobio Technology Company, Ltd., Beijing, China. FAW lures were installed in the traps at 0.5 to 0.7m and replaced every 20 days. Both dispensers and traps were installed as soon as the maize germinated after sowing, approximately one week after sowing. The sowing of all treatments at each location was completed within ten days to achieve a similar plant stage. Similar maize hybrids were planted in both pheromone treatment plots, pheromone foundational practice (PFP) and conventional grower’s practice (CGP), at each location to maintain uniformity.

Each trial location consisted of two treatments: pheromone foundational practice (PFP), treatment with dispensers containing blended FAW sex pheromone active ingredients, and conventional growers’ practices (CGP), treatment based on common growers’ practices to manage FAW. To ensure that the small plot size did not negatively affect the potential benefits of MD, the PFP treatment plots were 9-ha. The internal dose rate optimization of FAW dispensers tested with the same parameters as in these experiments, at ten dispenser increment levels from zero to 50 dispensers/ha revealed that a range of 20 to 40 dispensers/ha was an appropriate density (Provivi, internal data 2020). Since the minimum area treated is unnecessary for CGP or control, a 2-ha plot in the same area was demarcated. We reduced the edge effects wherever possible by selecting near-square-shaped PFP and CGP plots. A 200–500 m buffer was maintained between the PFP and CGP plots to minimize the drift of pheromones and contamination of treatments. All treatment plots were at least 100 m from a light source to avoid interference towards pheromone traps. A pheromone trap was installed in the middle of each hectare for each treatment. Additionally, to monitor the FAW population in the surrounding areas, four to six additional pheromone traps were installed in all directions of the trial location while maintaining at least 200 m from the PFP plots.

The PFP treatment plot was sprayed with approved insecticides at a labeled rate when a threshold of 20% of the damage was noticed, with damage defined as a Davis damage severity scale (Davis et al., 1992) of 3 out of 9 on the top three whorl leaves. For the CGP treatment, there was no restriction on insecticide application, and they were made according to each grower’s practice.

Moth capture. To understand trap reduction (MD), starting from the installation, moth capture assessments were made every seventeen days until harvest (~90 days crop). At each assessment, the male FAW moths caught in the traps were counted and the traps were reset with water mixed with non-odor detergent or oil.

Damage assessments. Vegetative damage assessments were conducted at regular intervals, starting seven days after the installation of dispensers in the field. Damage assessments were performed by scouting the damage severity caused by the FAW larvae. Damage in the vegetative stage was assessed on the top three whorl leaves based on the Davis damage severity scale of 0 to 9, where a higher severity denotes more significant damage by FAW larvae (Davis et al., 1992) Damage in the vegetative stages was assessed until the tasseling stage at seven-day intervals until 49 days after planting, with a damage rating of 20 plants per transect and 15 to 20 transects/ha. The distance between sampling points was 17 m.

Insecticide application. Insecticide application was recorded in both PFP and CGP treatment plots. The type of insecticide and dose rate were also recorded.

Trap reduction as a measure of mating disruption. The reduction in adult (male) moth trap counts in the PFP versus CGP is a proxy for mating disruption. Trapping reduction is the percentage decrease in moth captures in PFP traps compared to non-dispenser areas, such as CGP. Moth counts were modeled as an overdispersed Poisson variable as a function of pheromone treatment, time, and spatiotemporal random effects to account for the inherent correlations within and between trials. A hierarchical Bayesian framework was used to allow for the expected nonlinear, nonincreasing trend in trapping reduction over time. The nonlinear function used to model the trapping reduction over time is highly flexible and allows trapping reduction associated with the pheromone treatment to remain constant for many days before the expected decrease (at some point in time, all dispensers will eventually run out of pheromone active ingredient and will stop working), and then plateau at no pheromone effect, that is, no trapping reduction related to the pheromone treatments. Alternatively, if no trapping reduction is observed, the model allows this possibility. The model was fitted in MATLAB (Inc, 2021) using customized sampling algorithms.

Damage reduction as a result of mating disruption. This analysis aimed to determine whether there was a significant reduction in leaf damage in plots under the PFP treatment compared to plots under the CGP treatment. Damage reduction is the percentage decrease in the damage scale in PFP fields compared to non- dispenser areas such as CGP. The damage rating scale per sampling point was modeled as an over-dispersed ordinal variable as a function of pheromone treatment and spatiotemporal random effects to account for the inherent correlations within and between trials. The model was similar to a generalized linear mixed model (GLMM), but with spatiotemporal random effects defined as in the trap count model and not based on qualitative groupings. Again, a hierarchical Bayesian model framework was used, and the model was fitted through a customized slice-sampling algorithm in MATLAB (Inc, 2021) using customized sampling algorithms.

Insecticide usage reduction with mating disruption. The collected data from several insecticide spray applications in the PFP and CGP plots were analyzed by comparing them with a paired t-test in MATLAB (Inc, 2021).

Lower trap captures in the PFP than in the CGP indicate higher mating disruption. In the absence of pheromone dispensers, normal moth captures occur when pheromone traps are installed. FAW moths captured in CGP with no pheromones (control) in Indonesia’s wet season 2020 trial locations ranged from six to 30 months per trap. East Java had a higher moth pressure than South Sumatra Figure 1 Seasonal variation in the FAW moth pressure differed as moth pressure in control plots was higher (up to 60 moths/trap/day) in the northern and western parts of East Java (Situbundo and Nganjuk regencies) than in the southern part of the Jember regency (two to four moths/trap/day). Compared to Indonesia, four locations across the central and northern parts of Thailand captured fewer moths in the control plots, one to two moths/ trap/day Figure 2.

The analysis of trap reduction in Indonesia’s wet season 2020, as a measure of mating disruption, concluded that all PFP treatments, 20, 30, and 40 dispensers/ha, significantly reduced trap captures compared to control plots Figure 3 The analysis of moth captures in the five locations showed strong evidence of trapping reduction in all PFP treatments. The maximum trapping reduction was observed with 40 dispensers/ha, with a median trapping reduction of 88% (79–93% at 95% CI) over the entire 90-day season. However, 20 and 30 dispensers/ha treatments also provided season- long trapping reduction, with a median trapping reduction of 69% (47–86% at 95% CI) and 74% (60–85% at 95% CI), respectively Table 2.

The trapping results from the 2021 season, from seven locations in Indonesia and four locations in Thailand, showed a significant reduction in trapping in 30 and 40 dispensers/ha treatment plots Figure 3. A median trapping reduction of 90% (75–93% at 95% CI) and 93% (87–96% at 95% CI) was observed for 30 and 40 dispensers/ha treatments as season-long trapping reduction. The 20 dispensers/ha treatment also showed a season- long trapping reduction, with a median of 69% Table 2 However, wider credible intervals were observed in the 20 dispensers/ha (47–86% at 95% CI) treatment, showing less consistent disruption during the cropping season. Both the 30 and 40 dispensers/ha treatments led to significantly higher trapping reductions than the 20 dispensers/ha treatments (P < 0.05; Bayesian P-value).

Data analysis was performed on the damage caused by the larval population of fall armyworms among various pheromone treatments collected on the Davis Scale, 0 to 9, where a higher scale corresponds to more significant damage. Figure 4 shows the damage to the vegetative stages of maize (up to 45–50 days after planting) in the wet season of Indonesia in 2020. All five locations had an average score ranging between 0 and 5, whereas Southeast Asia 2021 had a higher range of average scores between 0 and 7 Figure 5 Two locations in Thailand (2021), Takhi, Nakhonsawan, and Sawang Arom, Uthaithani, experienced

more severe damage than the others. At various locations, the damage in the pheromone-treated plots was lower than that in the control plots. We also noticed that the 30 and 40 dispensers/ ha treatments had lower damage ratings than the 20 dispensers/ha and CGP treatments did. The severity of the damage, shown as the proportion of plants reaching various levels of the Davis Scale, is presented in Figure 6;Figure 7.

The severity of the damage, combined for five locations in Indonesia during the wet season of 2020, showed that the control plots, even when applied with more insecticide treatments, often reached the economic threshold of damage (20% of plants reaching more than three on the Davis Scale). However, the severity of damage in the pheromone treatments was lower, despite fewer insecticide sprays Figure 6 Damage severity at 11 locations across Indonesia and Thailand (2021) showed a similar trend, with a more significant proportion of plants reaching severe damage in the control plots Figure 7 We also noticed that the PFP 20 dispensers/ha treatment caused slightly more damage than the 30 and 40 dispensers/ha treatments.

Overall, damage reduction analysis from five locations in Indonesia during the wet season of 2020 showed that all PFP treatments (20, 30, and 40 dispensers/ha) worked very well, providing a median damage reduction of 35, 34, and 39% over the control plots, respectively Table 3 The differences between the PFP treatments were not statistically significant (P > 0.05; Bayesian P-value) Figure 8 For Southeast Asia’s 2021 damage results, all treatments significantly reduced the damage compared with the control

plots. The PFP 20, 30, and 40 dispensers/ha treatments provided a median damage reduction of 28, 35, and 39%, respectively, compared to the control plots. PFP 40 treatment led to significantly greater damage reduction than PFP 20 treatment (P < 0.05; Bayesian P-value) Table 3 However, the PFP 30 and 40 dispenser/ha treatments provided statistically equal damage reduction results.

Insecticide applications in all PFP plots were based on the threshold for the highest treatment rate of PFP 40, while control plots were sprayed at the discretion of the participating farmers. The number of insecticide applications in the PFP versus control plots is displayed for each location in Figure 9 At all locations, the pheromone treatment plots always received fewer insecticide sprays. At least two applications were made in the control plots and as many as eight applications were made in the vegetative phases of the crop. None of the PFP plots received more than two insecticide sprays per season to control FAW at all 16 locations. Overall, insecticide applications were analyzed by a paired t-test, which showed that PFP fields used significantly fewer insecticide applications (P < 0.001; from paired t-test) Figure 10.