How Reliable Are Decadal Climate Predictions of Near-Surface Air Temperature?

• DOI: 10.1175/jcli-d-20-0138.1
• Publication Date: 2020-11-02T14:23:39Z
• Authors: Deborah Verfaillie; Francisco J. Dobl...; Markus G. Donat; Núria Pérez-Zanón; Balakrishnan Sola...; Verónica Torralba; Simon Wild
• Keywords: Initialization; Forecast verification; Ensemble forecasting

Abstract Decadal climate predictions are being increasingly used by stakeholders interested in the evolution of over coming decade. However, investigating added value those initialized decadal other sources information typically generally relies on forecast accuracy, while probabilistic aspects, although crucial to users, often overlooked. In this study, quality near-surface air temperature from has been assessed terms reliability, an essential characteristic simulation ensembles, and compared reliability noninitialized simulations performed with same model ensembles. Here, is defined as capability obtain a true estimate uncertainty ensemble spread. We show limited initialization significantly more reliable than their counterparts only for specific regions first year. By analyzing different system we further highlight fact that combination models seems play important role size each individual system. This due sampling errors related physics, numerics, approaches involved multimodel, allowing certain level error compensation. Finally, study demonstrates all ensembles affected systematic biases dispersion affect reliability. set makes bias correction calibration necessary estimates probabilities can be useful stakeholders.