Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Skill assessment of seasonal hydroclimatic forecast for the region of Crete

Fragki Asimina

Full record

---

Summary

The need for weather forecasts is becoming increasingly important in modern times. From a utility point of view, in addition to the need for greater spatio-temporal detail, there is an increasing need to extend the forecast time horizon of our systems. This requirement is particularly strengthened in light of anthropogenic climate change with increased weather extremes such as storms, heat waves and water scarcity.The science of seasonal forecasting has a history of more than 30 years. Originating as a pure research activity by a few academic and research institutions, it has recently attracted considerable interest because of its potential to provide useful climate information at appropriate time horizons in a variety of areas, such as water resources management. Seasonal forecasts essentially contain a range of possible climate conditions over a forecast horizon of one to seven months. Although there is an abundance of data of this type, the degree of utilization is still low due to their low skill and the difficulty of processing them to transform them into useful information for decision support. This thesis aims to test the skill of seasonal climate forecasts and apply statistical processing for their improvement, focusing on the region of Crete. First, the predictability skill for one of the two most popular operational seasonal horizon forecasting systems, ECMWF System 5 (ECMWFS5) and Météo-France System 6 (MFS6), is evaluated. A set of methods is then applied for forecast calibration, bias correction, statistical downscaling, multivariate verification, and optimization to generate a tailored seasonal hydroclimate forecast for the region of Crete. The processing was performed using the Climate Services Toolbox (CSTools), a specialized R package that evaluates and improves the quality of seasonal and multi-year climate forecasts.Initially, a consistency check of the two systems was carried out and it was found that the best system for the area is ECMWFS5, which is also confirmed in the literature. Subsequently, the calibration application reduced the biases against the ERA5 reanalysis. Particularly for ECMWFS5 both temperature and precipitation show the smallest deviations compared to recorded values. Therefore, it was considered reliable for further processing. Downscaling was then applied as a fundamental step to increase the spatial resolution of the forecasts. Specifically, for the temperature variable, it was applied for the summer of 2003, as it was the warmest of the 1993-2016 period at the pan-European level, while for precipitation, it was applied for the autumn of 2002, the wettest of the corresponding period.Finally, the methodology was applied to develop focused probabilistic forecasts with an initial date of 1 October 2022 for the winter of 2023 and for the first two months of spring March-April. By evaluating the temperature results based on its increasing trend due to climate change, the probabilities for temperature and precipitation were largely realistic. The same applies to precipitation. However, a weakness of the forecasting systems was observed for the dry season, May to November, so there is room for improvements in predictability.Seasonal forecasts have great potential to provide valuable climate information that helps in decision-making for climate-related areas. The results confirm the potential usefulness of seasonal forecasting.

Available Files

Services