Abstract. A set of seasonal drought forecast models was assessed and verified for the Jaguaribe River in semiarid northeast Brazil. Meteorological seasonal forecasts were provided by the operational forecasting system used at FUNCEME (Ceará's research foundation for meteorology) and by the European Centre for Medium-Range Weather Forecasts (ECMWF). Three downscaling approaches were tested and combined with the models in hindcast mode for the period 1981 to 2014. The forecast issue time was January and the forecast period was January to June. Hydrological drought indices were obtained by fitting a 5 generalized linear model to observations. In short, it was possible to obtain forecasts for a) monthly precipitation, b) meteorological drought indices, and c) hydrological drought indices.The skill of the forecasting systems was evaluated with regard to root mean square error (RMSE) and the relative operating characteristic (ROC) skill score. Forecasts of monthly precipitation had little or no skill considering RMSE. Still, the forecast of extreme events of low monthly precipitation showed skill for the rainy season (ROC skill score of 0.24 to 0.33). A simi-10 lar picture was seen when forecasting meteorological drought indices: low skill regarding RMSE and significant skill when forecasting drought events of e.g. SP EI 01 (ROC skill score of 0.53 to 0.61). Similar results were obtained for low regional reservoir storage forecasts. Regarding the skill in the forecasted months, it was greater for April, when compared to February and March (the remaining months of the rainy season).This work showed that a multimodel ensemble can forecast drought events of time scales relevant to water managers in 15 northeast Brazil with skill. But no or little skill could be found in the forecasts of the whole range of monthly precipitation or drought indices (e.g. forecasting average years). Both this work and those here revisited showed that major steps forward are needed in forecasting the rainy season in northeast Brazil.

Abstract. The semiarid northeast of Brazil is one of the most densely populated dryland regions in the world and recurrently affected by severe droughts. Thus, reliable seasonal forecasts of streamflow and reservoir storage are of high value for water managers. Such forecasts can be generated by applying either hydrological models representing underlying processes or statistical relationships exploiting correlations among meteorological and hydrological variables. This work evaluates and compares the performances of seasonal reservoir storage forecasts derived by a process-based hydrological model and a statistical Driven by observations, both models achieve similar simulation accuracies. In a hindcast experiment, however, the accuracy of estimating regional reservoir storages was considerably lower using the process-based hydrological model, whereas the resolution and reliability of drought event predictions were similar by both approaches. Further investigations regarding the deficiencies of the process-based model revealed a significant influence of antecedent wetness conditions and a higher 10 sensitivity of model prediction performance towards rainfall forecast quality.Based on the findings of this study, we recommend using a statistical approach for merely predicting reservoir level and drought events at regionally and monthly aggregated scales. However, for forecasts at finer scales of space and time or for the investigation of underlying processes, the costly initialisation and application of a process-based model is worthwhile. Furthermore, the application of innovative data products, such as remote sensing data, and operational model correction methods, like 15 data assimilation, may allow for an enhanced exploitation of the advanced capabilities of process-based hydrological models.