SEASONAL STREAMFLOW FORECASTS IN AN ALPINE BASIN BY INTEGRATION OF REANALYSIS DATA AND SEASONAL CLIMATE FORECASTS

This research concentrated on assimilating the state-of-art ERA5 reanalysis dataset for snow and runoff modeling along with the System-5 ECMWF dataset for runoff forecasting in the European alpines situated in northern Italy. Even though the main focus of the study was to provide a suitable methodology for seasonal runoff forecast, initially, the suitability of the ERA5 dataset to model the hydrological parameters in the study basin was assessed. This assessment was a foremost requirement since the System-5 was forced with the initial condition provided by the ERA5 dataset and any errors in ERA5 translated into the forecasting variables as well. With the help of ICHYMOD hydrological, which has a built-in snow module, the preliminary study was able to quantify the errors in SWE followed by runoff modeling based on the errors of meteorological variables in the ERA5 dataset. Since the errors in ERA5 could have varying hydrological effects with respect to catchment spatial scale, 16 basins ranging from 10s to 1000s square Kms were considered. The precipitation of ERA5 was found to be suffering from overall bias originating from the lower precipitation events, whereas the temperature variable of ERA5 was displaying a colder bias. These errors in ERA5 simulated higher SWE variables in the high altitude smaller watershed with better runoff performance than the larger basins. However, with increasing basin size, the errors regarding the initial soil moisture condition get accumulated, due to bias in ERA5 lower precipitation. This error accumulation translates to poorer runoff performance for the larger basin. Interestingly, the ERA5 precipitation errors help to simulate comparable flood events for the larger basins owing to the higher initial moisture state before the flood and lower precipitation events during the flood. The correction in the ERA5 precipitation and temperature was achieved with quantile mapping. These corrections help substantially to remove the bias in ERA5 producing better SWE and runoff values across the basins. However, it is key to note the correction of temperature produced substantially better SWE results than the correction of precipitation, especially concerning smaller basins. The temperature correction, however, did not affect the runoff simulation and precipitation correction further depleted the flood simulation characterized by ERA5. Additionally, the ineffectiveness of ERA5 for the flood events was seen when the unitary basin size was considered. Corollary to the ERA5 precipitation, the system-5 also displayed a positive bias irrespective of the lead time. After the correction with quantile mapping considering the stations, the system-5 showed comparable precipitation values. Moreover, this correction assisted in a better runoff forecast as compared to the traditional Extended Streamflow Prediction (ESP) for lead time concerning up to 2 months. However, for a longer lead time, the ESP was comparatively better than the system-5 runoff forecasting method. In the case of baseflow forecasting, there was a gain of one-week lead time irrespective of the forecasting method. The system-5 forecasting of the sub-surface process did show some errors at a longer lead time, which could, firstly, be attributed to the difficulty in modeling alpine head basin characteristics of thin and highly variable soil variables and, secondly, to the inefficiency of the system-5 dataset to incorporate the chaotic meteorological information at longer lead times. Lastly, since the performance of the forecasting methodology depended on the initialization month, the role of the initial watershed state was found to be crucial for both the forecasting scheme.

Questa ricerca si è concentrata sull'assimilazione del set di dati di rianalisi ERA5 all'avanguardia per la modellazione di neve e deflusso insieme al set di dati ECMWF System-5 per la previsione del deflusso nelle Alpi europee situate nel nord Italia. Anche se l'obiettivo principale dello studio era fornire una metodologia adeguata per la previsione del deflusso stagionale, inizialmente è stata valutata l'idoneità del set di dati ERA5 a modellare i parametri idrologici nel bacino di studio. Questa valutazione era un requisito fondamentale poiché il System-5 è stato forzato con la condizione iniziale fornita dal set di dati ERA5 e qualsiasi errore in ERA5 si è tradotto anche nelle variabili di previsione.