Some specific scores use a neighborhood strategy in order to reduce double penalty effects, which penalize high-resolution models, compared to large-scale models. Contingency tables based on this strategy have already been proposed, but can sometimes display undesirable behavior. A new method of populating contingency tables is proposed: pairs of missed events and false alarms located in the same local neighborhood compensate in order to give pairs of hits and correct rejections. Local tables are summed up so as to provide the final table for the whole verification domain. It keeps track of the bias of the forecast when neighborhoods are taken into account. Moreover, the scores computed from this table depend on the distance between forecast and observed patterns. This method is applied to binary and multicategorical events in a simplified framework so as to present the method and to compare the new tables with previous neighborhood-based contingency tables. The new tables are then used for the verification of two models operational at Météo-France: AROME, a high-resolution model, and ARPEGE, a large-scale global model. The comparison of several contingency scores shows that the importance of the double penalty decreases more for AROME than for ARPEGE when the neighboring size increases. Scores designed for rare events are also applied to these neighborhood-based contingency tables.
The neighborhood-based ensemble evaluation using the Continuous Ranked Probability Score is based on the pooling of the Cumulative Density Function (CDF) for all the points inside a neighborhood. This methodology can be applied to the forecast CDF for measuring the predictive input of neighboring points in the center of the neighborhood. It can also be applied at the same time to forecast CDF and observed CDF so as to quantify the quality of the pooled ensemble forecast at the scale of the neighborhood. Fair versions of these two neighborhood scores are also defined in order to reduce their dependencies on the size of ensemble forecasts. The borderline case of deterministic forecasts is also explored so as to be able to compare them with ensemble forecasts. The information of these new scores is analyzed on idealized and real cases of rain accumulated during three hours and of 2-m temperature forecast by four deterministic and probabilistic forecasting systems operational at Météo-France.
RésuméAprès un rappel des principaux outils servant à la prévision du temps (essentiellement les modèles numériques simulant l'évolution de l'atmosphère), on décrit les critères et scores objectifs permettant de juger la qualité de la prévision. Ces critères varient beaucoup suivant l'échelle spatio-temporelle à laquelle on cherche à évaluer la prévision, l'échéance de la prévision (généralement présentée en mode déterministe à courte échéance, en mode probabiliste au-delà), mais aussi les caractéristiques plus ou moins extrêmes de l'événement que l'on cherche à prévoir. Une vérification régulière des prévisions permet aussi bien un suivi sur plusieurs décennies mettant en évidence les progrès à long terme qu'une évaluation détaillée des changements scientifiques et techniques effectués sur la chaîne de prévi-sion à une date donnée. AbstractThe verification of meteorological forecasts in Météo-France
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.