Premium
Simplifying and generalising Murphy's Brier score decomposition
Author(s) -
Siegert Stefan
Publication year - 2017
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1002/qj.2985
Subject(s) - brier score , decomposition , forecast verification , reliability (semiconductor) , simple (philosophy) , forecast skill , computer science , statistics , mathematics , ecology , power (physics) , physics , philosophy , epistemology , quantum mechanics , biology
The decomposition of the Brier score into Reliability, Resolution and Uncertainty has become a standard method in forecast verification. In this note a very simple derivation of the familiar Brier score decomposition is presented. The Reliability and Resolution terms can be calculated as average Brier score differences between the issued forecast, the recalibrated forecast and the climatological reference forecast. The result suggests a simple way to calculate similar decompositions for arbitrary verification scores, and that recalibration methods and reference forecasts can be chosen more flexibly than is generally appreciated. A new decomposition of the continuous ranked probability score (CRPS) is proposed.