Software library to quantify the value of forecasts for decision-making: Case study on sensitivity to damages

Abstract Decision making increasingly relies on forecasts of system behaviour. Relative Utility Value (RUV) is a new method to quantify the value of forecasts for decision making and can flexibly model many types of decisions. However, RUV is less accessible than other methods (e.g. Relative Economic Value) as it is relatively complex to implement and requires more inputs to use (e.g. explicit decision-context). This study introduces RUVPY, a reference implementation allowing easy specification of the decision context from included (or user-defined) decision rules, economic models, risk preferences, and damage functions. Capability is demonstrated through a case study on how damage functions affect streamflow forecast value in the Murray-Darling Basin, Australia. Forecast value is found to be highly sensitive to changes in the damage function. Implying that accurate estimates of value require the model of damages to match the real-world impacts. RUVPY can model realistic decision assumptions and provides researchers and practitioners with a versatile open-source tool for decision-making. Highlights • RUVPY software library now available to tailor assessment of forecast value to user decisions using Relative Utility Value (RUV). • Users can explicitly define the decision context from a set of decision types, economic models, decision rules, risk preferences, and damage functions. • Tailoring the damage function to match the damages of the real-world decision can be crucial to achieving reliable estimates of the true forecast value.