Calibration and Combination of CHFP precipitation forecast over South America using Ensemble Regression

CAROLINA VERA; OSMAN, MARISOL

Cape Town

Conferencia; IAPSO-IAMAS-IAGA Joint Assembly 2017; 2017

International Union of Geodesy and Geophysical

In this study we calibrate the climate predictionsavailable at the Climate Historical Forecast Project of the WorldClimate Research Program to develop seasonal precipitation forecasttools over South America. Forecasts made with initial conditions ofNovember and May and valid at December-January-February (DJF) andJune-July-August (JJA), respectively (Lead 1) over the period1982-2006 are considered. A Multi-Model Ensemble (MME) using 11models with around 10 ensemble members, was constructed and itsperformance was evaluated against CMAP database. The domain ofapplication spans [15°N-60°S;275°E-330°E] and it is also dividedin two subdomains: tropical South America and extratropical SouthAmerica.The Ensemble Regression technique (EREG) applies aregression equation developed for the ensemble mean to each ensemblemember to obtain a probability density function (PDF) whichrepresents the ensemble prediction. EREG is first applied to eachmodel and its ensemble members to calibrate them. Then, twoapproaches are used to obtain the consolidated PDF. The first oneconsists in using the weighted MME in a new ensemble regression,resulting in a weighted super-ensemble regression (WSEREG) to get theconsolidated PDF. The other technique consists in obtaining theconsolidated PDF computing the normalized summing of the weightedmodels? PDF (weighted kernels, WKERNELS). The consolidated PDFsobtained are used to forecast the three equally probable categoriesbelow, near and above normal. These forecast are confronted againstthose obtained counting the proportion of ensemble members of the MMEfalling in each category (counting estimate technique, CE).Results show that both WKERNELS and WSEREGoutperform CE in terms of the Ranked Probabilistic Skill Score (RPSS)and Brier Skill Score (BSS) in both seasons. However, only innorthern South America the performance of both consolidationtechniques is slightly better than the climatological values of thepredictand (three categories equally probable). In extratropicalSouth America both RPSS and BSS values change from less than -0.5 forCE to near 0 for both WKERNELS and WSEREG.On the other hand,reliability diagrams computedover the entire domain shows thatWKERNELS and WSEREG substantially improve the forecast in terms ofreliability respect to than obtained with CE.p { margin-bottom: 0.1in; direction: ltr; color: rgb(0, 0, 0); line-height: 120%; }p.western { font-family: "Liberation Serif", "Times New Roman", serif; font-size: 12pt; }p.cjk { font-family: "Noto Sans CJK SC Regular"; font-size: 12pt; }p.ctl { font-family: "FreeSans"; font-size: 12pt; }