Uncertainty of heat wave metric calculations

Climate/weather extremes such as heat waves (HWs) are of the great interest to study as they have the significant harmful effect on the environment and society. There are many researches dealing with the calculation of HW metrics and their long-term trends on both the global and regional/national spatial scale. In our work based on a case study of Ukraine, we aimed to quantify the uncertainty of HW metric calculations, which might originate from climate input data. To this end, we used a mini statistical ensemble of several gridded data sets of maximum daily air temperature (TX), covering the territory of Ukraine for the period of 1950-2020 (70 years) with the same spatial resolution. The ensemble included ERA5 reanalysis data (remapped by means of the CDO software to the finer grid of 0.1^ox0.1^o with different interpolation algorithms), ERA5-Land, E-OBS (the ensemble mean) and Ukrainian gridded observation data previously developed for the period of 1946-2020. We defined a HW as an event when conditions (TX in our case) above criteria (90-th percentile calculated based on the WMO standard 1961-1990 reference period) persist at least three consecutive days, with permission of a 1-day time gap. Four HW metrics were considered, namely heat wave number (HWN), duration (HWD), frequency (HWF) and amplitude (HWA). The calculation of yearly time series of the HW metrics was performed by means of the R package heatwaveR for each grid point of the domain and each member of the constructed statistical ensemble. The uncertainty of the HW metrics was defined as a difference between min and max metric’s values calculated for different members of the ensemble. We also calculated the range of the possible variations in long term trends of obtained yearly time series of the HW metrics. Our results showed that depending on climate data used for HW climatology analysis, the calculation results might differ significantly for a particular grid point and year. However, on average (over the whole domain and the period under study), variation of the HW metrics is not so pronounced. Moderate variations are also observed in long-term trends of the metric time series.

 

This work has received funding through the MSCA4Ukraine project, which is funded by the European Union