Inter-model variability in the influence of El Niño-Southern Oscillation over the precipitation in South America

Global Climate Models (GCMs) are projecting future changes in different modes of variability influencing precipitation over South America (SA). Among them, El Niño-Southern Oscillation (ENSO) is the most important mode of inter-annual variability. However, the GCMs still present high inter-model variability. Therefore, it is still necessary to advance in the assessment of simulated ENSO impact before asserting future consequences. The aim of this study is to analyse inter-model variability in ENSO signal on precipitation simulated by GCM-CMIP6 over SA.

Daily precipitation and monthly sea surface temperature (SST) were obtained from 15 GCMs selected from CMIP6 and compared to ERA5 reanalysis, for the period 1981-2010. The ENSO was characterised through the Oceanic Niño Index (ONI) which was calculated based on SST anomalies over the Niño3.4 region. Total accumulated rainfall (PRCPTOT) was calculated in two trimesters October-December (OND) and December-February (DJF). These seasons were chosen because of the incidence of ENSO signal over SA.

Firstly, inter-model variability in the ONI values was assessed comparing the distributions with the index obtained from ERA5 and quantifying the number of cases under each ENSO phase: El Niño, La Niña and Neutral. The inter-quartil range is underestimated by 53% of the models and overestimated by one model, for both seasons. The rest of the models present similar distribution to ERA5. Consequently, the models that underestimate the inter-quartil range, overestimate the number of Neutral cases. Additionally, the extreme values of El Niño phase are more overestimated than the values of La Niña phase.

Secondly, the simulation of ENSO signal on PRCPTOT was assessed through Spearman correlation (5% significance level) and composite patterns. The analysis was focused on two main regions where ONI signal is stronger: Northern South America (NSA) and Southeastern of South America (SESA).

In general terms, for OND, the models are able to capture spatial patterns, in particular, with positive correlations over SESA and negative ones over NSA with 70% inter-model agreement. The rest of the models present higher spatial variability. The ensemble of the models also captures the spatial pattern correctly in almost all South America.

The ENSO signal in PRCPTOT for DJF is weaker, according to ERA5. The ensemble of the models captures the sign of the signal over the regions of interest, but fails over central Brazil, located among SESA and NSA. The level of agreement between the models is similar to OND over the regions with strong ENSO signal but, over transitional regions, the inter-model variability is higher.

Based on these results, composite analysis was carried out for the ensemble of the models. In general terms, the signal simulated by the GCMs is weaker than ERA5, but they adequately identify the regions and the sign of the signal.

The main result of this research is that ENSO signal on South America precipitation is well simulated by GCMs particularly over the regions where this signal is stronger. This study is a first step for a subsequent analysis of the future projections of the ensemble of the GCMs, considering other precipitation indices.