Development of statistical models for at-site probabilistic seasonal rainfall forecast

Authors

  • Gabriele Villarini,

    Corresponding author
    1. Department of Civil and Environmental Engineering, Princeton University, Princeton, NJ, USA
    2. Willis Research Network, London, UK
    • Department of Civil and Environmental Engineering, Princeton University, Princeton, NJ, USA.
    Search for more papers by this author
  • Francesco Serinaldi

    1. Willis Research Network, London, UK
    2. School of Civil Engineering and Geosciences, Newcastle University, Newcastle Upon Tyne, UK
    Search for more papers by this author

Abstract

A probabilistic seasonal rainfall forecasting system for the Bucharest-Filaret (Romania) station based on Generalized Additive Models in Location, Scale and Shape (GAMLSS) is proposed. First we develop statistical models to describe seasonal rainfall over the period 1926-2000, both considering the seasonal record as a continuous time series and accounting for seasonal changes, and by developing ad hoc models for each individual season. The Southern Oscillation Index (SOI), the North Atlantic Oscillation (NAO) and the seasonal rainfall for the previous year are included as possible covariates. Model selection is performed with respect to two penalty criteria [Akaike Information Criterion (AIC) and Schwarz Bayesian Criterion (SBC)], each of them leading to different final model configurations in terms of predictors and their functional relation to the parameters of the probability distribution. Retrospective forecast, in which the parameters of the models are re-estimated every time new information becomes available, is performed on a yearly basis for the period 1986-2000. The quality of the forecasts is assessed in terms of several accuracy measures and by visual examination of the forecasts' probability distributions. The best forecasts are obtained for the winter season. While it is not possible to identify a single ‘best’ model according to all the forecast measures, we recommend using the model that considers the seasonal rainfall as a continuous time series and penalized with respect to AIC. Copyright © 2011 Royal Meteorological Society

Ancillary