Characteristics of droughts in CMIP6 models

Abstract

The goal of this Master thesis is to evaluate the representation of drought events in historical simulations carried out with the new generation of Earth system models (ESMs) participating in the Coupled Model Intercomparison Project Phase 6 (CMIP6), and to characterize how frequency, intensity and duration of droughts are changing in the future under a specific forcing scenario (here SSP5-8.5). Results from different drought indices, in particular the Standardized Precipitation Index (SPI) that is based on precipitation only and the Standardized Precipitation Evapotranspiration Index (SPEI) are compared. For SPEI, different methods for calculating potential evapotranspiration (PET) have been proposed in the literature following the equations of Thornthwaite (SPEI-Th), Hargreaves (SPEI-Hg), and Penman Monteith (SPEI-PM). These methods approximate PET with increasing complexity and number of variables: SPEI-Th is based on temperature only, SPEI-Hg uses monthly mean daily minimum and maximum temperatureand SPEI-PM is additionally taking into account cloud cover, air pressure and wind speed. Due to the lack of available output variables and the verification of all SPEI methods for the historical period, for the analysis of ESMs often the methods based on empirical correlations between temperature and PET have been applied. However, using output from three CMIP6 ESMs that provided all variables required for the calculation of all three SPEI methods plus the internally simulated PET variable, this thesis in a first step shows that only the SPEI-PM method is able to correctly reproduce drought characteristics similar to the results achieved if the model’s PET variable is used. While SPEI-PM projects a moderate increase in extreme drought events especially in the subtropics, SPEI-Th projects an unrealistic drastic increase in the frequency and duration of droughts nearly globally and SPEI-Hg a strong increase in subtropical regions and the mid-latitudes, both inconsistent with calculations based on the models’ PET variable. In a second step, a larger ensemble of seven CMIP6 models is therefore only analyzed with the SPI and SPEI-PM indices. Except for certain regions, SPI and SPEI-PM drought characteristics calculated from the multi-model mean in comparison to ERA-Interim and the Climate Research Unit (CRU) dataset for the period 1979 to 2014 only show small deviations. Percentage changes in drought characteristics between the future (2050-2100) and the historical (1950-2000) period for SPI and SPEI-PM show in general good agreement, although SPEI-PM results in larger changes in drought frequency compared to SPI. SPEI-PM shows an increase in frequency and severity of drought events in the already dry subtropics by over 200% and 50% while SPI shows an increase in same characteristics by over 150% and 30%, respectively.