Statistical prediction of Winter rainfall in Patagonia (Argentina)

GONZÁLEZ, MARCELA HEBE; HERRERA, NATALIA

Horizons in Earth Science Research

nova publisher

Lugar: nueva york; Año: 2014; p. 221 - 238

This chapterdescribes rainfall evolution in southern Argentina (Patagonia) and faces with the possibility to predict winter rainfall using oceanic and atmospheric predictors. The Andes mountainranges along the west of Patagonia and a large plateau extends towards the east.A rainfall trend analysis was derived;a decrease of rainfall was observed in all seasons in central and east Patagonia meanwhile an increment is observed in southern part of Los Andes. The study area was regionalized using a principal component analysis and four regions were defined. Almost all regions showed an annual cycle with maximum rainfall in Winter. Therefore, several large scale oceanic and atmospheric forcings (sea surface temperature, geopotential height at low, middle and high levels and wind at low levels) in Aprilwere used to evaluate their influence on winter (May to July, MJJ) Patagonia rainfall using different statistical models: an autoregressive integrated moving average model (ARIMA), Holt Winter (HW), Climate Prediction Tool (CPT) and an ensamble of all, called multimodel. ARIMA model predicts future values of a time series by a linear combination of its past values and a series of performing a maximum likelihood fit of the specified ARIMA model to the time series. HW model is based on exponential smoothing. It is a technique which uses a set of exponential weights that decrease each time by a constant ratio giving more weight to the more recent data values and less weight to the data values from further in the past. All data values in a series contribute to the calculation of the prediction model.Climate Predictability Tool (CPT), developed by The International Research Institute for Climate and Society (IRI, Columbia University, cpt@iri.columbia.edu) provides a computational package for constructing a seasonal climate forecast model, using a set of predictors which must be carefully selected.All the models were used to predict seasonal rainfall in winter (MJJ) in the four regions defined in Patagonia. They were designed with MJJ rainfall for the period 1975-2007 in CPT model and for the period 1988-2007 in ARIMA and Holt Winter because this last two models use the previous values to generate the final scheme. All the models were proved for 2008-2012. The CPT model uses oceanic and atmospheric predictors and therefore it contains the influence of oceanic and atmospheric mechanisms associated with rainfall, meanwhile ARIMA and HW uses only the values of rainfall series and thus they face with the memory of the data. The main result of this workis that the consideration of ARIMA and HW in the multimodel improves the efficiency obtained using CPT. This result was supported by the calculation of efficiency coefficients which represent the ability to detect over and sub normal rainfall. Additionally, the probability functions resulting from estimated and observed rainfall were similar at the 95% confidence level and revealed that multimodel provides reasonably good forecasts.