Abstract 116

Abstract ID: 116

Skill assessment of weekly temperature anomalies in the SubX Project for the extended austral summer in South America

Lead Author: Lucia Micaela Castro
Universidad de Buenos Aires, Facultad de Ciencias Exactas y Naturales, Departamento de Ciencias de la Atmósfera y los Océanos – Servicio Meteorologico Nacional, Argentina

Keywords: SubX, Subseasonal, South-America, Predictability, Summer

Abstract: The predictability of six models participating in the Subseasonal eXperiment project (SubX) and the corresponding Multi-Model Ensemble (MME) are evaluated for extended austral summer (October-April) 2-meter temperature (T2M) weekly-mean anomalies in South America. Hindcasts for the 1999-2014 period are used to produce weekly means for weeks 1 to 4 (spanning lead times from 1 to 28 days). For simplicity only the MME is addressed. The role of the Madden-Julian Oscillation (MJO) in the predictability levels is also analyzed. To study extreme events predictability, weeks with T2M anomalies above (below) the 90th (10th) percentile are evaluated as well.

In week 1 most of the continent presents 50% to 60% of predictability, while the north of Chile and eastern Brazil have values from 60 to 70%. Meanwhile, northwestern Brazil, Colombia and northern Peru have lower predictability, ranging from 30 to 40%. By week 2, land predictability drops to 10-20% in the south, 40-50% in the northeast, and 30-40% everywhere else. Weeks 3 and 4 are similar, with values below 30% for most of South America. It is found that tropical latitudes, north of 30°S, retain predictability for longer lead times than higher latitudes.

For MME initializations during active MJO events, predictability of weeks 2-3 can increase, with respect to inactive, in southeastern South America (SESA) up to 20%, and up to 40% when restricting only to initializations during phases 4-5. It is especially beneficial for Patagonia, in week 3 improving 30% for forecasts initialized in phases 2-3 and 20% in phases 4-5.

Model verification shows for week 1 a 0.7 of Anomaly Correlation Coefficient (ACC) south of 20°S, and 0.5 in the north. In week 2 the ACC drops to 0.5 and by weeks 3-4, the southern part has an ACC of 0.3, while losing significance in central South America. In week 1, the Root Mean Square Error (RMSE) north of 20°S is 1.0°C and 1.5°C in the south. For the southern area, the RMSE increases along lead time reaching values of 3.0°C in week 4; while in the northern side ACC and RMSE remains the same as week 1.

As observed for predictability, verification skill scores are also favored by the MJO. In particular, two spots show benefit of this condition: in week 3, Patagonia has lower errors for all MJO phases, and an ACC increment for forecasts initialized in phases 4-5. Secondly, SESA, shows smaller errors in week 3 for all MJO phases and better correlations in phases 1, 4, 5 and 8. Alternatively, in week 3 MJO is unfavorable to central South America, rising errors in all phases and reducing ACC for every phase except 6 and 7.

Predictability for extreme events above (below) 90th (10th) percentile is calculated. It is found that extreme warm cases are better predicted than normal in Colombia, Peru and Venezuela for all lead weeks. Furthermore, cold cases are predictable in Patagonia in week 2, and in most parts of Argentina in weeks 3-4. For week 3, forecasts show performance improvements in eastern Brazil for both extreme cases.

Finally, the 2013 December heat wave, characterized by a positive T2M anomaly moving northwards across southern South America and a negative anomaly in eastern Brazil, is evaluated. Out of the 6 SubX models, 4 are able to forecast this event 15-21 days in advance, though two failed to capture its magnitude (CCSM4, GEM). Also, 5 of the models can forecast both anomalies 8-14 days before. FIMr1p1 stands out because it reproduces the anomaly movement with 15-21 leads. MME is not able to capture this extreme event.

Marisol Osman: Universidad de Buenos Aires, Facultad de Ciencias Exactas y Naturales, Departamento de Ciencias de la Atmósfera y los Océanos. CONICET. Universidad de Buenos Aires. Centro de Investigaciones del Mar y la Atmósfera (CIMA). CNRS, IRD, CONICET .UBA. Instituto Franco-Argentino para el Estudio del Clima y sus Impactos (IRL 3351 IFAECI). Now at Karlsruhe Institute of Technology –
Mariano Alvarez: Independent consultant