Assessment of an Ensemble of Artificial Neural Networks Seasonal Precipitation Forecasting Model based on Climate Oscillation Indices. A Case Study of Ceará, northeastern Brazil.

Código

I-EBHE0052

Autores

Enzo Pinheiro, Taha B.M.J. Ouarda

Tema

WG 2.3: Near-term (annual to decadal) forecasts of water availability

Resumo

This research assesses the skill of an ensemble of Artificial Neural Networks (EANN) for 1-month lead seasonal precipitation forecasting. The EANN employs October-November-December low-frequency climate oscillation indices to predict February-March-April total precipitation anomalies in the Brazilian state of Ceará, a prominent region for climate forecasting studies. The EANN members are derived using the Bagging algorithm. The model's deterministic and probabilistic forecasting ability is compared to traditional statistical models, i.e., Multiple Linear Regression and Multinomial Logistic Regression, and to dynamical models that compose Ceará?s operational seasonal forecasting system. The forecast verification is carried out through a leave-one-out cross-validation based on 40 years of data. Additionally, the study proposes combining the EANN with dynamical models into a hybrid multi-model ensemble (MME). A spatial comparison indicates that the EANN was among the models with the lowest Root Mean Squared Error and Raked Probability Score, except in the southern region of the state where the climate oscillation signal is limited. An analysis of the area-aggregated reliability and sharpness diagrams shows that the EANN is better calibrated than the individual dynamical models and has better resolution than traditional statistical models for above-normal (AN) and below-normal (BN) categories. Both statistical and dynamical models depict a bad-calibrated near-normal (NN) category. Finally, it is also shown that combining the EANN and dynamical models into a hybrid MME reduces the overconfidence of the extreme categories observed in a dynamically-based MME, improving the reliability and resolution of the forecasting system.