Spatial Dynamics and Hidden Spread of Banana Bunchy Top Disease in Benin

Merveille Koissi SAVI^1*Corneille Ahanhanzo²Fidele Tiendrebeogo¹Angela O. Eni¹Justin S. Pita¹

• ¹Félix Houphouët-Boigny University, Abidjan, Côte d'Ivoire
• ²Universite d'Abomey-Calavi, Cotonou, Benin

Banana Bunchy Top Disease (BBTD), caused by the Banana Bunchy Top Virus (BBTV) and transmitted by the aphid Pentalonia nigronervosa, poses a growing threat to banana production in West Africa, often resulting in 100% yield loss in infected plantations. Yet its epidemiological dynamics remain poorly known. Current management in Benin relies on visual symptom identification and informal seed networks, both of which are vulnerable to the pathogen's prolonged latency. We conducted a structured cross-sectional survey of 176 banana farms across 12 departments of Benin between December 2024 and February 2025, complemented with archival surveillance data from 2018–2020. Apparent disease incidence was estimated from visual inspection and corrected for diagnostic error using a hierarchical Bayesian misclassification model. Extreme gradient boosting (XGBoost) identified wind speed in April, sucker density, and September maximum temperature as the primary drivers of symptom expression (AUC = 0.913). Bayesian adjustment for imperfect sensitivity (Se ≈ 0.78) and specificity (Sp ≈ 0.92) revealed that true incidence exceeded field estimates by a median factor of 2.1 (95% CrI 1.6–2.8), exposing substantial under-detection of infection in southern agroecological zones. Integration of bias-adjusted posterior incidence across years reconstructed the epidemic wavefront, indicating a northward expansion from Akpro-Missérété (6.6° N) to ~ 9.8° N by 2025. Linear regression of front displacement on time yielded a mean spread rate of 37.8 km yr⁻¹, with residual patterns suggesting acceleration during 2020–2022, likely due to secondary introductions or intensification of local transmission. This study provides the first spatially explicit quantification of BBTD spread in Benin, demonstrating that visual field assessments substantially underestimate the true burden of the disease. The integration of Bayesian bias correction and wavefront modeling provides a robust framework for mapping and forecasting the spread of plant diseases under imperfect detection conditions.