Development of a high-resolution database for daily precipitation in Greece

The accurate assessment of precipitation is a critical challenge in meteorology due to the non-normal distribution characteristics commonly associated with precipitation data. This distribution can lead to significant errors in forecasting models, particularly concerning extreme precipitation events, which are both infrequent and increasingly influenced by climate change. The implications of climate change on the frequency and intensity of these extreme events further complicate the task of accurate prediction, necessitating improved methodologies for rainfall estimation.

In the context of Greece, the challenge is intensified by a sparse network of precipitation observation stations. This limited data availability, coupled with the region's inherent geographical variability—characterized by diverse topographic features such as mountains and valleys—creates additional hurdles in the generation of reliable precipitation datasets. Consequently, the objective of this study is not only to address these challenges but also to create a high-resolution precipitation database specifically for Greece, employing advanced statistical interpolation techniques.

To achieve this, we systematically investigate a range of interpolation techniques aimed at generating high-resolution gridded daily precipitation datasets across the Greek territory. Our approach utilizes a comprehensive dataset of meteorological stations, which forms the backbone of our analysis. In addition, we incorporate geographical variables derived from satellite-based elevation data and integrate precipitation data sourced from the ERA5 atmospheric reanalysis, a product known for its high spatial and temporal resolution.

Three distinct modeling approaches are developed throughout this research.

• General Additive Model and Indicator Kriging: In the first approach, we employ a General Additive Model combined with an Indicator Kriging methodology, relying predominantly on the station data and a limited selection of geographical variables. This foundational model serves as the baseline for understanding the initial relationships between observed precipitation and geographical factors.
• Incorporation of ERA5 Data: The second iteration enriches the interpolation methodology by blending ERA5 reanalysis data with the observational datasets. In this stage, we expand the geographical variables used, allowing for a more nuanced understanding of precipitation patterns in relation to the diverse topography of Greece.
• Multi-Model Interpolation Framework: Lastly, we introduce a novel modeling framework that not only integrates ERA5 data and an array of geographical datasets but also employs a multi-model interpolation process. This strategic approach utilizes different models tailored to predict precipitation during distinct thresholds. By addressing various precipitation intensity levels, this model enhances the ability to accurately forecast both average and extreme precipitation events.

The results of this study demonstrate that the inclusion of ERA5 data can significantly enhance the accuracy of the interpolated precipitation, particularly in regions where the observational station dataset is sparse. Moreover, the implementation of multi-model interpolation techniques—where distinct models are utilized for different precipitation thresholds—offers substantial improvements in the accuracy of both total precipitation forecasting and the modeling of extreme precipitation events. This multi-faceted approach effectively addresses crucial limitations exhibited in previous modeling efforts, thereby contributing valuable insights and robust methodologies to the field of meteorological research in Greece.