Quantifying CMIP6 Model Uncertainty in Extreme Precipitation Forecasts – Is it up because of that?

From extreme weather and climate

AmalJohn ^{one c} HerveDouville ^one AurélienRibes ^one PascalYiou ^b

^one Center National de Recherches Météorologiques, Météo-France, CNRS, Toulouse, France
^b Laboratoire des Sciences du Climat et de l’Enosystemnement, UMR 8212 CEA-CNRS-UVSQ, IPSL & U Paris-Saclay, 91191 Gif-sur-Yvette, France
^c Université de Toulouse, France

Received October 8, 2021, Revised February 9, 2022, Accepted March 10, 2022, Available Online March 21, 2022, Version of the Record March 26, 2022.

Https: //doi.org/10.1016/j.wace.2022.100435 Get rights and content
By Creative Commons license Open access

abstract

Expected rainfall change extremes and their uncertainties are assessed using a set of global climate models from period 6 of the Paired model comparison project (CMIP). They are scaled by the respective changes in global mean surface temperature (ΔGSAT) or local surface temperature (ΔT) and expressed as a 20-year return value (RV20) of precipitation. up to one day per year. Our main objective was to quantify the model response uncertainty and highlight regions where the changes may not be consistent with the widely used assumption of the Clausius-Clapeyron (CC) ratio. is ≈7%/K. When using a single perception for each model, as in the latest report of the Intergovernmental Panel on Climate Change (IPCC), the spread between models is estimated. The price includes both model uncertainty and internal variability, however can be assessed separately using a large population. Although the difference between models is overestimated, our results show a strong enhancement of extreme precipitation with more than 90% of the models simulating an increase in RV20. Furthermore, this increase is consistent with a CC ratio of ≈7%/K over approximately 83% of global land area when scaled (ΔGSAT). Our results also support the generation of multiple clusters of single-model baseline conditions in subsequent CMIP forecasts, to better filter internal variability when estimating responses to extreme events.

1. Introduction

Global climate models provide an increasingly comprehensive representation of the climate system and are used as a key tool for understanding and forecasting changes in climate mean, variability, and poles due to human activities. The Intergovernmental Panel on Climate Change (IPCC) in its sixth assessment report (AR6) re-estimated the observed increase in mean global surface temperature of 1.09°C over the period. period 2011–2020 compared to the beginning of the industrial revolution (1850–1900 ), which can be entirely attributed to human influence (IPCC AR6 SPM Masson-Delmotte et al. (In 2021)). This anthropogenic global warming is thought to have long-term consequences for all components of the climate system, including changes in the distribution of daily precipitation. Several generations of multi-model simulations contribute to Paired model comparison project (CMIP), supported by observational evidence, shows that both the frequency and intensity of extreme daily precipitation events have increased in recent decades (Allen and Ingram, 2002, Asadieh and Krakauer, 2015, Scherrer et al., 2016, Karl and Easterling, 1999, Kharin et al., 2013, Min et al., 2011, O’Gorman, 2015). This is also documented in the IPCC special report on Risk management of extreme events to promote adaptation to climate change (SREX, Seneviratne (2012)).

In the absence of moisture limitation and significant kinetic response, the rain intensity is expected to increase exponentially with atmospheric temperature at a rate determined by the Clausius-Clapeyron (CC) relationship. A strong ratio of daily extreme precipitation to global warming under the scenarios confirmed by Li et al. (2020) who found that? rainfall change extremes follow the changes in global warming with a CC ratio of approximately ≈7%/°C in the latest generation CMIP6 samples. Several studies based on climate modeling simulations show an increase in future extreme precipitation with temperature at a rate equal to or higher than the CC rate (Li et al., 2020, Kharin et al., 2007, Pall et al., 2007, Allan and Soden, 2008, Sugiyama et al., 2010, Kao and Ganguly, 2011, Muller et al., 2011). However, wet extremes are not expected to increase in all regions (Trenberth, 2011, Pfahl et al., 2017).

All of these studies indicate the mean or median of the multiple models and have not yet assessed the uncertainty in the global CMIP6 projections. A different set of model projections often presents a large disparity (Lehner et al, 2020) and may even disagree on whether a particular area becomes wetter or drier (a sign of future change). Even if there is general agreement among models about the signs of changes at the extremes expected by a warmer climate, the magnitude of those changes can vary considerably. Although climate models have improved in recent decades (Wyser et al, 2020, Zelinka et al, 2020), these improvements do not necessarily lead to a reduction in the variance between the forecasts (Douville et al., 2021). Therefore, the main focus of this paper is to quantify the model uncertainty in extreme precipitation forecasts based on the CMIP6 model. We also aim to provide a blueprint for using these projections to identify areas where projected changes in daily precipitation are consistent with CC rates, and areas where area no.

Changes in precipitation extremes in the context of a warming climate arise from both thermodynamic and dynamic effects (Pfahl et al., 2017). A subtropical relationship or even a negative dependence on global mean temperature has been found for precipitation extremes in some regions, especially for oceanic regions dry in the subtropics, probably as a result of reduced moisture availability and large-scale intensification subsidence (Berg et al., 2009, Hardwick Jones et al., 2010, Utsumi et al., 2011, Pfahl et al., 2017). But the question of the appropriate choice of temperature for mitigating extreme precipitation remains an open question, and existing studies vary in extent (Zhang et al, 2019, Schroeer and Kirchengast, 2018, Sun et al, 2021). There is a large-scale warming contrast between continental land and oceans with certain regions of the ocean having negligible or limited variation in expected surface temperatures. The greater warming observed over land is likely to result in a lower ratio to the local mean temperature, which may not be considered a sub-CC expansion rate (Wang et al., 2017). Any difference from the CC rate could be indicative of a dynamic response that could be amplified or offset by the area thermodynamic response (Pfahl et al., 2017, Sherwood et al., 2010, O’Gorman, 2015). So, here we explore changes in extreme precipitation that are proportional to changes in the global or local mean surface air temperature.

Learn many times (Alexander et al., 2006, Tebaldi et al., 2006, Sillmann et al., 2013a, Sillmann et al., 2013b) used different indices to represent different characteristics of precipitation extremes. Here we focus on extreme events with a typical 20-year payback (or 20-year payback value, RV20) estimated from annual maximum one-day precipitation (RX1DAY). The changes in RX1DAY return value over the long run are expected to be larger than the changes in the average RX1DAY return value and increase with increasing rarity (Mizuta and Endo, 2020, Wehner, 2020). Here, we did not explore longer return periods (e.g. 50 or 100 years) because the associated measurement uncertainty is even stronger than our RV20 estimate due to the limited sampling.

The objective of this study was to evaluate the uncertainty of expected changes in precipitation extremes based on the CMIP6 multi-model assembly, to discuss the limitations of assessing the spread between models using such chance combinations and highlighting areas where expected changes may not be consistent with the widely used assumption of a Clausius-Clapeyron ratio of 7% /K (Kharin et al., 2013, Westra et al., 2013, Seneviratne et al, 2021). For this purpose, we use the SSP5-8.5 scenario from 35 CMIP6 models. Thus, the total variance in this set is a combination of both model response uncertainty and internal variability. Therefore, we also evaluated the potential contribution of internal variation to inter-model spread by analyzing the expected changes of RV20 in the CanESM5 model with 25 runs. , with the same focus script.

The rest of the paper is structured as follows. We start by introducing in Section 2 models and methods used in this study. Jump to results in Section 3, we address the uncertainties in the model’s forecasts along with a discussion of the role of internal variability using synthetic simulations from CanESM5. The role of local versus global temperature scales was also assessed. Section 4 summarize the main findings. Other supporting figures and tables are available in the online supplemental documentation.

Read the full article here.