Can we forecast farmers' yields? The relationships between rainfall variability, farmers' expectations, and actual yields in a tropical dryland

Climate variability is one of the most important drivers of global crop yield variability. This is particularly true in dryland ecosystems, where vegetation dynamics, including agricultural production, are strongly shaped by both the pronounced seasonality of rainfall and its intra-seasonal distribution, both of which can exhibit substantial variability. This variability plays a key role not only in impacting end of season yields, but is also likely to influence farmers' expectations of final yields and thus their management decisions. Both these values are typically collected using farmer surveys, and their values influence higher-level policy decisions related to food security. It is unclear, however, whether farmers' expectations are shaped by the rainfall variables that most influence final yields, which has important ramifications for food security. In this study, we use agricultural survey data combined with a bias-corrected gridded meteorological forcing dataset to investigate the relationships between farmers' expected yields, end-of-season yields, and several different indices of seasonal rainfall and rainfall variability. We focus on Zambia, a country that relies heavily on smallholder agricultural production for the majority of its food production. Our goals are to identify which aspects of rainfall variability have the greatest influence on farmers' yield expectations, which have the greatest impact on actual yield, how well actual and expected yields are correlated with one another, and how these relationship differ throughout the country. Understanding these connections can help to improve seasonal crop yield forecasting and farmer decision-making, and thereby lead to improved food security policy in a region where rainfall variability is increasing.