The performance of scientific, indigenous, and hybrid weather forecasts: Systematic evaluation and farmers’ perception in Bangladesh

Access to reliable and skillful weather information could assist smallholder farmers in Bangladesh to reduce their vulnerability to rainfall variability and extremes. The available weather forecast information, however, is still limited in the provision of daily location-specific weather information for smallholder farmers in Bangladesh. Because of this reason, the use of local forecasts is a more favorable and affordable way of accessing weather information for many farmers in low-latitude developing countries. In the WATERAPPscale project, we have initiated a climate information service that provides timely and location-specific weather forecasts for smallholders and established 16 farmer’s weather schools across Bangladesh, where training on interpretation of scientific forecasts (SF), collection of local forecasts (LF) data, and sharing forecast information took place. This study aims to systematically evaluate the performance of the SF and LF used by farmers, by applying a dichotomous method to distinguish yes/no rainfall events. The results were compared with farmers’ perception of the forecast skills. In addition, the skill of a simple hybrid forecast (HF), which is an integrated system of SF and LF, was assessed. The SF and LF data were obtained from the meteoblue hindcast and from the questionnaires, respectively. This study used ERA5 and ground observation datasets as benchmarks for the weather forecasts. Results show that overall, the LF has slightly higher skill than the SF when compared to the ERA5 dataset. The forecast performance, however, reduces by almost half when the ground-based observation is used instead of ERA5, associated with a high false alarm. The evaluation results, however, are contradictory to the farmers’ perception that SF has a much higher performance than LF. Combining the SF and LF into a simple HF generates higher skill than any single forecast alone, which highlights the necessity to develop a hybrid forecast that combines scientific and indigenous weather forecasting for farm decision-making. The developed HF system will deliver a reliable forecast, trustworthy, and conserved indigenous knowledge that has been passed down from generation to generation.