Uncertainties in projections of climate extremes indices in South America via Bayesian inference

Abstract

Historical simulations and projections of climate extremes indices of precipitation and temperature were analysed over South America until the end of the 21st century through 31 general circulation models (GCMs) under four Representative Concentration Pathways. Simulations were compared with reanalysis data, and a Bayesian inference method was used to assess the uncertainties involved in the multi-model climate projections. Regarding the precipitation extremes indices, the GCMs' simulations reasonably approached the reanalysis data, but with heterogeneous biases, both in sign and in the location of the highest values. The temperature extremes indices presented the smallest biases when compared to precipitation. Projections show a gradual growth of precipitation extremes events as the analysed radiative forcing scenario increases, both in magnitude and extent, over a large part of South America. Projections also indicate a decrease in cold days and nights and an increase in warm days and nights, more pronounced in the equatorial region. Bayesian inference method smoothed changes in precipitation extremes events, both in magnitude and extent, compared to the simple GCMs' ensemble mean. There was no considerable variation in the temperature indices when applying the Bayesian inference. Finally, the probability density functions resulted in a predominance of multimodal and wide curves for the precipitation indices, showing great uncertainties in the GCMs' results, differently from those for the temperature indices, where the GCMs presented good agreement represented through unimodal and narrow curves.

Current and Future Climate Extremes Over Latin America and Caribbean: Assessing Earth System Models from High Resolution Model Intercomparison Project (HighResMIP)

“…Overall, these results indicate warmer conditions during the day and night, an increase in the duration of warm episodes over most of South America, and consistently a decrease in cold days and nights, which is in line with the others (Collazo et al 2022 ; Gouveia et al 2022 ). In general, the MMEs are adequate in representing the observed trends for each index in the regions of South America on an annual time scale, except for DTR, where the most remarkable inconsistencies are present.…”

“…Other works assess the skills of the different gridded datasets (e.g., ESMs, reanalyses, satellites products) to reproduce the spatial–temporal variability (Avila-Diaz et al 2021 ; de Lima and Alcântara 2019 ; Kim et al 2020 ; Ongoma et al 2018 ). In this sense, we evaluated the performance of CMIP6 models in simulating the ETCCDI indices using annual summary values, similar to other studies (Aerenson et al 2018 ; Gouveia et al 2022 ; Thibeault and Seth 2014 ) that indicate that most of the impactful climate extremes can be described by annual indices. Additional details about the performance of CMIP6 models over North, Central, and South America on the annual cycle of total precipitation and mean temperature are discussed by Almazroui et al ( 2021a , b ).…”

“…Providing reliable information regarding historical and future projections of such events represents an enormous challenge for climate researchers (McPhillips et al 2018 ; Medeiros and Oliveira 2022 ; Mistry 2019 ; Mysiak et al 2018 ; Santos et al 2017 ). Previous studies have investigated the historical evolution of climate extremes in different parts of the world, including Latin America and the Caribbean, using the set of indices established by the Expert Team on Climate Change Detection and Indices (ETCCDI) (Aguilar et al 2005 ; Avila-Diaz et al 2020b ; Cornes and Jones 2013 ; Donat et al 2013 , 2016 ; Dunn et al 2020 ; Gouveia et al 2022 ; Kitoh and Endo 2016 ; Nakaegawa et al 2014 ; Skansi et al 2013 ; Zilli et al 2017 ). Those studies have indicated two components linked to uncertainties of gridded or observational datasets (e.g., reanalyses and satellite products) and climate simulations (e.g., Earth system models—ESMs) concerning estimates of climate extremes events at local and regional scales: (i) these events have large temporal and spatial variability (Akinsanola et al 2020 ; Avila-Diaz et al 2020b ; Campozano et al 2016 ; Na et al 2020 ); (ii) the assessments of climate extremes in the Northern Hemisphere are more abundant and reliable when compared to the Southern Hemisphere (Donat et al 2016 ; Lehmann et al 2015 ; Sillmann et al 2013 ), due to the fact that in the latter there is a greater lack of climatic data for an analysis of long periods and low spatial distribution of meteorological stations (Condom et al 2020 ; Liebmann and Allured 2006 ; Pabón-Caicedo et al 2020 ; Solman 2013 ).…”

Assessment of dry and heavy rainfall days and their projected changes over Northeast Brazil in Coupled Model Intercomparison Project Phase 6 models

“…For example, Gouveia et al . (2022) using the ensemble mean of 31 GCMs from CMIP5 showed that the number of consecutive dry days (CDD) may increase at 35 days over most of the NEB in the end of the 21st century. Similarly, positive change (>75 mm) for very wet days (R95p) are also projected.…”

“…Moreover, results obtained through climate change projections indicate that the frequency and severity of these extreme events is expected to continue to increase in the future due to global warming (IPCC, 2013;Sillmann et al, 2013;Marengo et al, 2020;IPCC, 2021;Medeiros et al, 2022). For example, Gouveia et al (2022) using the ensemble mean of 31 GCMs from CMIP5 showed that the number of consecutive dry days (CDD) may increase at 35 days over most of the NEB in the end of the 21st century. Similarly, positive change (>75 mm) for very wet days (R95p) are also projected.…”