Local knowledge on flood forecasting and meteorological phenomenon have received scant research attention in South Asia. While local communities are often recognised as key producers of knowledge on adaptation and resilience, their role as producers of knowledge of meteorology and flood forecasting has drawn lesser interest than their consumption of it as end users. Moreover, the opinion that has carried through to limited research on this issue has been largely that of men. This paper attempts to address such a research gap, by recording various sophisticated means deployed by local communities living in villages in transboundary Gandak River basin in Bihar, India, to forecast floods and heavy rainfall. This research documents the gendered process by which local knowledge is produced through complex interaction between fine-grained observations and official early warning systems. It also explores how communities practice knowledge innovation by making generic and centralized flood forecasting information locally applicable, through triangulation. Our research argues that ‘local’ should not be recognised purely for dissemination of flood early warning information but also as a place of knowledge generation on flood forecasting. Current discourse on flood forecasting needs to recognize the gendered production of meteorological and flood forecasting knowledge in local communities. Furthermore, strengthening local knowledge systems on flood forecasting can work to counterbalance the drawbacks of centralized flood early warning systems, provided gender concerns embedded in both are recognised. Based on field research, this paper concludes that production and consumption of flood forecasting knowledge needs local and scientific communities to work together for reducing knowledge gaps at both ends.