A methodology for cross-timescale, calibrated and flexible forecasts

Abstract

Successful climate services involve the use of tailored regional forecasts at single or multiple timescales. The implementation of such forecasts is not always straightforward, and depends on several different factors, like which variables, models and calibration methods to use, how to create the ensemble and tailoring, or even how to present them to the decision makers. In this talk I’ll discuss NextGen, IRI's systematic, general approach for designing, calibrating, combining, and verifying objective forecasts, aligned with WMO’s recommendations for objective forecasts. NextGen involves the identification of decision-relevant variables with the stakeholders, and the analysis of the physical mechanisms, sources of predictability and suitable candidate predictors for those key relevant variables, in observations and models. If prediction skill is deemed actionable, NextGen helps select the best (dynamical/statistical/hybrid) models for the region of interest through a physical-process-based evaluation and automates the generation and verification of tailored multi-model, locally- or pattern-based-calibrated predictions at multiple temporal and spatial scales. The system takes advantage of the expertise of local scientists and decision-makers in the implementation countries to maximize predictive skill and the potential to tailor the forecasts. Several concrete examples will be discussed, from “usual” climate forecasts to applications for forecast-base financing and predictions of coffee and rice yield, energy, acute under-nutrition cases, potential risk of transmission of mosquito-borne diseases, and human migration.