Five Lessons for Closing the Last Mile: How to Make Climate Decision Support Actionable

Abstract

Climate shocks are increasing, threatening global agricultural production and food security. But a more extreme climate allows for improved predictions and enables advisory services that allow farmers, ranchers and consumers to respond effectively. To date, there is limited uptake of forecasts. How can we make sure these predictions are valued by and valuable for users of agro-climatic forecasts? Over the past two years, we held over 40 interviews with food system stakeholders to identify their needs and shortcomings of existing decision support. In this Commentary, we combine these findings and nascent modeling efforts with existing literature to characterize five lessons for improving the uptake and utilization of predictive tools for last mile users in the agrifood system. Given the explosion of machine learning prediction efforts across many applications, we believe our lessons are broadly applicable to forecasting models intended for decision support. Improved accuracy alone does not necessarily lead to improved decision support, and the trust required to motivate action.