On the assessment of temperature and precipitation extremes in Central Europe in gridded observation datasets and reanalysis

As the climate continues to warm, patterns of air temperature and precipitation in Central Europe (CE) are changing, affecting the frequency, intensity, and spatial extent of extremes in this region. Studying and understanding these variations is crucial due to natural hazards such events trigger and is essential for refining climate predictions. However, the reliability of climate reanalysis or gridded datasets regarding extreme weather is still uncertain at varying temporal and spatial scales, especially in areas of diversified geography like CE. In this study we aim to evaluate the performance of two reanalysis, ERA5 and ERA5-Land, and two gridded observation datasets, EOBS and newly developed high-resolution data for CE, alongside in-situ observations, regarding the occurrence, spatial extent, and intensity of extreme temperature and precipitation events. The study is based on daily maximum and minimum temperatures (Tmax, Tmin), and precipitation totals (RR) for the 1961-2023 period, covering CE defined as the area encompassing Poland, Czechia, Slovakia, and eastern Germany. The extreme events are defined based on absolute as well as local percentile thresholds. The newly introduced daily gridded dataset for CE is based on quality-controlled and homogenised station data from approx. 500 climatological stations and approx. 1000 rain gauges, with a spatial resolution of 1×1 km and additionally 0.5×0.5 km in the mountainous areas to better capture details of complex terrain.

Gridded datasets validation procedure comprises of the comparison of in-situ data with the results extracted from the nearest grid points as well as from the predefined areal buffer. The co-occurrence of extremes between the datasets will be examined using statistics based on a contingency table, such as the hit rate, while the other extreme events’ characteristics, i.e. their spatial extent and intensity, will be studied employing the most common statistical measures, i.e. correlation coefficients, bias estimation, cross validation.

The results indicate that the data accuracy regarding the representation of the temperature and precipitation extremes in CE depends strongly on the subregional topography features as well as varies with the intensity and severity of the events. These conclusions highlight the necessity of local datasets development, alongside with already existing regional or global databases, especially when specialised environmental modelling is taken into consideration.