Abstract

Extreme precipitation during critical harvest periods is a major source of agricultural losses and damage for smallholder farmers, particularly for crops that are highly sensitive to untimely rainfall. This research investigates how voice-based climate advisory services can support farmer decision-making under rainfall uncertainty during mung bean (Vigna radiata) harvesting in Bangladesh. Using a participatory mixed-methods approach, we co-designed, implemented, and evaluated an automated voice alert system that delivered location-specific five-day rainfall forecasts and harvest advisories directly to farmers' mobile phones. The research combined farmer co-design, rainfall monitoring, advisory dissemination, and post-season telephone surveys to examine farmers' access to, comprehension of, and self-reported responses to advisory messages, as well as their perceived economic value. Farmer responses varied according to the timing of rainfall events relative to harvesting periods and crop maturity. Responses were higher when heavy rainfall occurred during active harvesting periods and crops had reached harvestable maturity, but lower when harvesting had already been completed or crops were still immature. Across multiple seasons, farmers reported perceived avoided losses valued at USD 299-333 ha-1 yr-1. Extrapolation of these farmer-reported estimates to the population of advisory recipients who reported taking protective actions suggested cumulative avoided losses of approximately 1235-3407 t yr-1. The estimated annual value of farmer-reported avoided losses ranged from USD 0.55-1.38 million. These findings suggest that the usefulness of climate advisory services depends not only on forecast information but also on the alignment between weather risks, crop development stages, and available management options.