This research was conducted in 50 localities from the cotton-producing region of Côte d’Ivoire, which extends from the central to the northern part of the country between latitudes 7˚5 to 12˚N and longitudes 3˚ to 8.5˚W (Figure 1). This region spans both sides of the 9th parallel. The area north of the 9th parallel is characterized by a Sudanian climate featuring an extended dry season from October of one year to May of the subsequent year. Conversely, the southern area situated below the 9^th parallel experiences greater precipitation and is characterized by a sub-Sudanian climate with two distinct rainy and dry seasons. The rainy season occurs from March to July and from October to November, whereas the dry season occurs from August to September and from December to February. The annual rainfall in this region varies between 1300 and 1700 mm [10]. Fifty localities were selected from the entire cotton-growing region.

Surveys were conducted during four consecutive cotton-growing seasons (2021-2024). For each season, field observations were carried out from July to November, covering the main vegetative and reproductive stages of the crop. Each year, 50 localities were monitored across the cotton-producing region of Côte d’Ivoire (i.e., the same 50 localities were followed over the four years). In each locality, 10 farmers’ cotton fields were selected (500 fields per year). When a selected farmer did not cultivate cotton each year, a replacement field was selected within the same locality and within a radius of ≤1 km to maintain spatial comparability.

Within each locality, the 10 selected fields were distributed to capture local variability while remaining feasible for weekly monitoring. Two criteria guided selection: (i) spatial coverage within the locality (North-South and East-West directions), and (ii) sowing period. To limit spatial autocorrelation, sampled fields within a locality were separated by at least 300 m whenever possible (and never less than 100 m), as determined by road distance or GPS tracks during field visits. Sowing dates were recorded for all selected farmers, grouped into 10-day intervals (decades), and fields were then sampled across these sowing decades proportionally to the number of farmers who sowed within each decade, followed by a random draw within each decade.

To ensure valid comparisons among years, the same monitoring protocol was applied in each season (2021-2024). The same 50 localities were followed annually, with the same target sampling intensity (10 fields per locality per year). In every field, observations were conducted within a standardized 0.25 ha central area using identical plant- and boll-inspection procedures, the same observation frequency (weekly plant inspections and five boll dissections per field), and the same observation window (approximately 30 - 122 days after sowing). All monitored fields were managed under the standard national cotton protection program (six insecticide sprays at 14 day intervals), thereby reducing interannual variability due to differences in control practices.

In each selected field, a 0.25 ha usable area was delimited at the plot centre to reduce edge effects and standardize observations across visits and years. The cotton variety cultivated in monitored fields was Gouassou Fus [11]. The crop was protected according to the standard national program recommended to farmers in Côte d’Ivoire, consisting of six insecticide applications at 14 day intervals from approximately 45 to 115 days after emergence.

In addition to pest and damage observations, the following variables were recorded for each field and visit: locality name, field identity (farmer/plot), sowing date (used to classify sowing period), and sampling date (expressed as days after sowing). This survey did not directly measure climatic variables (e.g., temperature, rainfall, humidity) at the field level, nor did it quantify cotton acreage by region. Therefore, statements regarding climatic drivers or regional changes in planted areas are treated as contextual interpretation and are supported by published sources where cited rather than inferred from the present dataset.

Source: Base map data from: National Bureau of Technical Studies and Development (BNETD). 2014.Updated using ArcGIS 10.6.

Figure 1. Study area.

2.3.1. Counting of Bollworms on Cotton Plants

Weekly direct observations were conducted from 30 to 122 days after sowing. At each visit, 30 cotton plants were inspected within the 0.25 ha observation area. Plants were selected as six groups of five consecutive plants along a diagonal transect of the field [12], as shown in Figure 2, to ensure representativeness. The position of the diagonal was changed each week to improve field coverage across the season.

Each plant was meticulously examined to count D.watersi, H. armigera, and Earias spp. larvae. These species are classified as exocarpic because the larvae inhabit and develop external bolls on the cotton plant.

Figure 2.Method for selecting six groups of five plants along the diagonal.

2.3.2. Sanitary Analysis of Green Bolls

The larvae of P.gossypiella and T.leucotreta develop within bolls and are thus categorized as endocarpic. Consequently, a sanitary analysis of bolls is imperative for their detection. Starting on the 80^th to 108^th day after sowing, five health evaluations were conducted weekly. During each evaluation, 50 green bolls were randomly collected within the 0.25 ha plot, with a maximum of one boll per plant. However, the cotton plants in the four central rows were spared as they were reserved for yield estimation. These bolls, which were approximately the same age, were sampled from the middle of the plant and close to the main stem. After collecting, the bolls were examined and categorized into several classifications: healthy, perforated bolls, and bolls that appeared healthy but had internal damage.

During the sanitary analysis, larvae of P.gossypiella and T.leucotreta found inside bolls were counted.

2.3.3. Harvest and Estimation of Cotton Yield

Harvesting was performed in the four central rows of each plot when bolls were fully open. Seed cotton was weighed after harvest, and yield was expressed in kilograms per hectare.

Analyses were performed to assess temporal (year) and spatial (locality/production area) variability in bollworm abundance, and to quantify relationships between pest pressure, boll damage and seed cotton yield.

Because larval counts are discrete and repeatedly collected, larval abundance data were analysed using generalized linear mixed models (GLMMs), with a Poisson or negative binomial distribution depending on dispersion. To account for the hierarchical sampling design and repeated measurements, the random effects Locality and Field were included in the models to evaluate their influence on pest dynamics. A linear regression analysis was performed between the percentage of damaged bolls and seed cotton yield to identify the relationships between these two variables.

Statistical analyses were carried out using GenStat (10^th edition), and distribution maps were produced using ArcGIS 10.6.

Overall mean densities of the five bollworm taxa recorded during 2021-2024 are summarized in Table 1. For exocarpic larvae, the mean numbers (±SD) per 30 plants were 0.038 ± 0.011 for D.watersi, 0.14 ± 0.061 for H. armigera, and 0.06 ± 0.012 for Earias spp. For endocarpic larvae, mean densities per 100 green bolls were 0.014 ± 0.003 for P.gossypiella and 0.001 ± 0.00 for T.leucotreta.

Table 1. Average densities of bollworms on plants and in cotton bolls.

3.2.1. Variation of Exocarpic Bollworms

Annual mean densities of exocarpic bollworms differed significantly for all three taxa (Table 2). For D.watersi, mean density increased from 0.03 ± 0.00 larvae per 30 plants in 2021 to a maximum of 0.07 ± 0.01 in 2024. For H. armigera, density decreased in 2022 (0.06 ± 0.03) relative to 2021 (0.11 ± 0.04), peaked in 2023 (0.22 ± 0.09), and then declined in 2024 (0.13 ± 0.05). For Earias spp., density was lowest in 2022 (0.05 ± 0.01) and was higher in 2021 and 2024 (0.07 ± 0.02 and 0.07 ± 0.01, respectively), with intermediate values in 2023 (0.06 ± 0.00).

Table 2.Trends in the average densities of bollworms on cotton plants and within green bolls from 2021 to 2024.

3.2.2. Variation of Endocarpic Bollworms

For P.gossypiella, mean density varied among years, with the highest value observed in 2023 (0.02 larvae per 100 bolls) and the lowest values in 2021 and 2022 (0.006 and 0.009 larvae per 100 bolls, respectively). Overall, densities showed an increasing tendency from 2021 to 2024. In contrast, T.leucotreta was consistently rare, with mean densities close to zero in all years and no significant interannual variation.

3.3.1. Impact of the Production Area on Exocarpic Bollworms

Figure 3 shows notable variations in the distribution of D.watersi across different locations. The highest larval densities were recorded in the northeast, specifically in Ferké (0.46 larvae per 30 plants), Koumba (0.22 larvae per 30 plants), and Sikolo (0.20 larvae per 30 plants). Conversely, densities approaching 0 larvae per 30 plants were observed in Bouandougou, Boundiali, Boron, Niofoin, Odiéné, Sarhala, Sinématiali, Tienko, Tomono, Vavoua and Zuénoula. The comprehensive analysis presented in Table 3 indicates that this species predominantly inhabits northern cotton-growing areas.

Figure 4 illustrates the distribution of H. armigera across production areas. This species occurred in all production areas, with the highest mean density recorded in Sordi (0.81 larvae per 30 plants), followed by Napié (0.47 larvae per 30 plants) and Diawala (0.41 larvae per 30 plants). In contrast, lower densities were observed in localities such as Boundiali, Ouéllé, Goulia and Gbon, with means close to 0.00 larvae per 30 plants. The comparative analysis in Table 3 indicates a significant difference between the northern and southern parts of the cotton-growing region, with higher densities in the north (0.16 larvae per 30 plants) than in the south (0.10 larvae per 30 plants).

Significant spatial variation in Earias spp. densities was observed (Figure 5), with the highest mean densities in the northeastern region. The largest mean density was recorded in Diawala (0.45 larvae per 30 plants), followed by Ferké and Sikolo (0.24 larvae per 30 plants each). Conversely, very low densities (approximately 0.00 larvae per 30 plants) were recorded in Boundiali, Bouandougou and Sinématiali. The analysis in Table 3 confirms that mean densities of Earias spp. were higher in the northern region than in the southern region.

Source: Figures 3-7 are all sourced from field survey data (primary data).

Figure 3. Spatial variation in the mean density of Diparopsiswatersi larvae within cotton-growing areas of Côte d’Ivoire based on surveys from 2021 to 2024.

Figure 4. Spatial variation in the mean density of Helicoverpa armigera larvae within cotton-growing areas of Côte d’Ivoire based on surveys from 2021 to 2024.

Figure 5. Spatial variation in the mean density of Earias spp. larvae within cotton-growing areas of Côte d’Ivoire based on surveys from 2021 to 2024.

3.3.2. Impact of the Production Area on Endocarpic Bollworms

Figure 6 illustrates the notable variation in larval densities of P.gossypiella across different localities. This species is predominantly observed in southern regions, specifically in the localities of Zuénoula (0.22 larvae/100 bolls), Kounahiri (0.21 larvae/100 bolls). Vavoua (0.14 larvae/100 bolls) and Koumbala (0.06 larvae/100 bolls). The analytical results presented in Table 3 suggest a higher abundance of these species in the southern cotton-growing basin.

Figure 6.Spatial variation in the mean density of Pectinophoragossypiella larvae within the cotton cultivation areas of Côte d’Ivoire based on a survey from 2021 to 2024.

Figure 7 depicts the overall low density of T.leucotreta. The highest densities were recorded in Boundiali (0.05 larvae/100 bolls) and Mankono (0.02 larvae/100 bolls). This species did not exhibit a distinct distribution pattern between the northern and southern regions (Table 3).

Figure 7.Spatial variation in the mean density of Thaumatotibialeucotreta larvae within cotton-growing areas of Côte d’Ivoire based on surveys from 2021 to 2024.

Table 3.Comparison of bollworm densities between the northern and southern production areas of the cotton-growing region.

Table 4summarizes bollworm (fruit-feeding Lepidoptera) densities according to sowing period.

No statistically significant differences were detected in the average densities of Diparopsiswatersi, Earias spp., P.gossypiella, and T.leucotreta across the various sowing periods. In contrast, a significant variation in the density of H. armigera was observed depending on the sowing period. The highest density (0.18 larvae per 30 plants) was recorded during the D4 sowing period (June 21 to 30), whereas the lowest density was noted during the D1 sowing period (May 20 to 31). Overall, there was an observable trend indicating a progressive increase in H. armigera density from the early to the late sowing periods.

Table 4.Variation in the densities of bollworms in relation to the timing of cotton sowing in Côte d’Ivoire.

3.5.1. Influence on the Rate of Perforated Bolls

Table 5presents the multiple regression model relating the mean number of perforated bolls to larval densities. The model was significant (p < 0.001) and indicated that Lepidopteran larvae explained approximately 10.5% of the variation in boll perforation (R² = 0.105). Significant predictors were D.watersi, H. armigera, Earias spp. and P.gossypiella. All regression coefficients (β) were positive, indicating increased boll perforation with increasing larval density. Standardized coefficients suggested that D.watersi had the strongest association with boll perforation, followed by P.gossypiella and H. armigera.

3.5.2. Influence on the Rate of Internal Damage in Cotton Bolls

Table 6 presents the regression analysis of internally damaged bolls in relation to larval densities. The model was significant (p < 0.001) and indicated that approximately 3% of the variation in internal boll damage was explained by bollworm larval densities (R² = 0.03). Coefficients (β) were positive for P.gossypiella and T.leucotreta, indicating higher internal damage with increasing densities of these endocarpic larvae. In contrast, coefficients were negative for D.watersi, H. armigera and Earias spp., consistent with their primarily external feeding on bolls. Standardized coefficients suggested that D.watersi, P.gossypiella, and T.leucotreta contributed most to the model.

Figure 8illustrates a significant negative correlation (r = -0.45) between seed cotton yield and the percentage of damaged green bolls. Approximately 20% of yield losses are related to damage caused by bollworms (R² = 0.20).

Table 5.Regression coefficients of the rate of perforated bolls as a function of bollworm densities (F = 29.62, p < 0.001).

R² = 0.105; SE: standard error; 95% CI: confidence interval; β std: standardized coefficients.

Table 6.Regression coefficients for the rate of bolls with internal damage as a function of bollworm densities (F = 9.94, p < 0.001).

R² = 0.03; SE: standard error; 95% CI: confidence interval; β std: standardized coefficients.

Figure 8.Variation in seed cotton yield in response to bollworm damage on green bolls.