Attributing trends in extremely hot days to changes in atmospheric dynamics

This paper presents a method for attributing regional trends in the frequency of extremely hot days (EHDs) to changes in the frequency of the atmospheric patterns that characterize such extraordinary events. The study is applied to mainland Spain and the Balearic Islands for the extended summers of the period 1958–2008, where significant and positive trends in maximum temperature (<i>T</i>_<i>x</i>) have been reported during the second half of the past century. <br><br> First, the study area was split into eight regions attending to their different temporal variability of the daily <i>T</i>_<i>x</i> series obtained from the Spain02 gridded data set using a clustering procedure. Second, the large-scale atmospheric situations causing EHDs are defined by circulation types (CTs). The obtainment of the CTs differs from the majority of CT classifications proposed in the literature. It is based on regional series and on a previous characterization of the main atmospheric situations obtained using only some days classified as extremes in the different regions. Three different atmospheric fields (SLP, T850, and Z500) from ECMWF reanalysis and analysis data and combinations of them (SLP–T850, SLP–Z500, and T850–Z500) are used to produce six different CT classifications. Subsequently, links between EHD occurrence in the different regions and CT for all days have been established. Finally, a simple model to relate the trends in EHDs for each region to the changes in the CT frequency appearance has been formulated. <br><br> Most regions present positive and significant trends in the occurrence of EHDs. The CT classifications using two variables perform better. In particular, SLP–T850 is the best for characterizing the atmospheric situations leading to EHD occurrences for most of the regions. Only a small number of CTs have significant trends in their frequency and are associated with high efficiency causing EHD occurrences in most regions simultaneously, especially in the northern and central regions. Attribution results show that changes in circulation can only explain some part of the regional EHD trends. The percentage of the trend attributable to changes in atmospheric dynamics varies from 15 to 50 %, depends on the region and is sensitive to the selected large-scale variables