| ABSTRACT | The spring streamflow in the Iberian Peninsula (IP) is forecasted by using the climatic variables sea surface temperature (SST), geopotential high (Z500), air temperature (TMP) and precipitation (RR) as predictor fields. After an exhaustive quality control of data, the compiled dataset of streamflow comprises 325 gauge stations in Spain, 9 in Portugal, and 170 reservoir entrances in Spain, totalling 504 data series, covering the period from October 1975 to September 2008. Firstly, the significant modes of covariability between these potential predictors and the spring streamflow are identified using lagging singular value decomposition (SVD) technique. The SVD was performed considering the atmospheric and oceanic fields leading the spring streamflow for lags ranging from four seasons (4S scenario) to one season (1S scenario). Secondly, an evaluation of the predictors stability is carried out analysing the stability of the correlation between the predictors and the spring streamflow at each gauge station through a running window correlation approach. The forecasting model is obtained by means of a stepwise regression between the stable predictors and the spring streamflow. In order to avoid artificial skill effects in forecasting, a leave-one-out cross validation was performed. A set of verification measures has been used: Pearson’s correlation coefficient, the Root Mean Square Error Skill Score and the Gerrity Skill Score. Results show that there is not any predictor for 4S scenario and the quality of number of stations forecasted increase as lag between predictors and predictands decrease. In the 3S scenario, a total of 230 stations are forecasted, with some skill identified in 77 of them, mainly located in the northern half of IP, in particular in the Miño-Sil and Douro Basins. In the scenario 1S, the number of stations forecasted increases up to 267. This increase is more remarkable in the Mediterranean slope and in the Tagus Basin. Acknowledgements: This work has been financed by the projects P11-RNM-7941 (Junta de AndalucÃa-Spain) and CGL2013-48539-R (MINECO-Spain, FEDER). |