reforecasting

Reforecasting is the practice of updating forecasts as new information becomes available, often by applying improved models, data, or methods to past or current forecast problems. The term is used across disciplines, including meteorology, climate science, and finance, to describe the act of revising predictions in light of better data or methods.

In meteorology and climate science, reforecasting refers to the systematic reconstruction of past forecasts using a consistent numerical weather prediction system and data assimilation setup. A reforecast dataset is produced by repeatedly running the forecast model for historical dates with the improved model, observations, and initialization. The goal is to create a long, homogeneous archive that enables accurate verification, bias correction, and reliable climate and trend analyses. Reforecast datasets support skill assessment of forecast systems and help calibrate probabilistic forecasts.

The typical workflow involves collecting historical observations, updating initial conditions with current assimilation techniques, running forecasts from those dates through a standardized model configuration, and applying post-processing or statistical calibration. Because the process is computationally intensive, it is usually conducted by national meteorological agencies or research centers for selected time spans or key variables.

In business and economics, reforecasting describes revising forecasts for metrics such as revenue, demand, or costs when new information becomes available. This may employ time-series updating, Bayesian methods, or machine learning to produce revised projections and inform decision-making. Reforecasting aims to improve accuracy and responsiveness across forecasting contexts.