Comparing Temperature and Precipitation Extremes Across Multiple Reanalyses and Gridded in Situ Observational Datasets

Wednesday, 17 December 2014: 11:35 AM
Markus Donat1, Lisa V. Alexander1, Jana Sillmann2, Simon Wild3, Francis W Zwiers4 and Tanya Lippmann1, (1)University of New South Wales, Climate Change Research Centre, Sydney, Australia, (2)University of Oslo, Oslo, Norway, (3)University of Birmingham, School of Geography, Earth and Environmental Sciences, Birmingham, United Kingdom, (4)University of Victoria, Vancouver, BC, Canada
Changes in climate extremes are often monitored using global gridded datasets of climate extremes based on in situ observations or reanalysis data. This study assesses the consistency of temperature and precipitation extremes between these datasets. We compare temporal evolution and spatial patterns of annual climate extremes indices across multiple global gridded datasets of in situ observations and reanalyses to make inferences on the robustness of the obtained results.

While there are distinct differences in the actual values of extremes, normalized time series generally compare well and temporal correlations are high for temperature extremes, in particular for the most recent three decades when satellite data are available for assimilation.

Extreme precipitation is characterized by higher temporal and spatial variability than extreme temperatures, and there is less agreement between different datasets than for temperature. However, reasonable agreement between gridded precipitation extremes from the different datasets remains.

While there is general agreement between the different reanalyses and gridded observational data in regions with dense observational coverage, different reanalyses show trends of partly opposing signs in areas where in situ observations are sparse, e.g. over parts of Africa and tropical South America. However, in the absence of reliable observations it is difficult to assess which reanalyses are more realistic here than others.

Using data from the 20th Century reanalysis and a novel century-long gridded dataset of extremes we also investigate consistency of extremes from these two datasets back to the beginning of the 20th Century. Global average time series of different extremes indices compare generally well over the past 70 years but show larger differences before around 1940. However, in areas with good observational coverage, including North America, Europe and Australia, agreement remains strong also throughout the earlier decades of the 20th Century.