Steady global surface warming from 1973 to 2022 but increased warming rate after 1990

Global mean surface temperature (GMST) is a crucial indicator of climate change. However, year-to-year fluctuations due to internal variability obscure long-term trends and make assessing recent warming rates challenging. The years 2020-2022 featured prolonged La Niña conditions, which exert a cooling effect on global temperatures, potentially masking the true rate of anthropogenic warming. Simultaneously, the balance of anthropogenic radiative forcing is also dynamic. Methane concentrations have risen sharply recently, while CO2 emission increases have slowed. Furthermore, the decrease in SO2 emissions (due to air pollution control efforts) has implications for aerosol-climate interactions and near-term temperature evolution. Understanding the interplay of these factors requires careful analysis to separate internal climate variability from the underlying anthropogenic warming signal. This study builds on a previously published methodology to filter out the influence of sea-surface temperature (SST) patterns on GMST, allowing for a more robust assessment of the warming rate, particularly in recent years affected by La Niña.

Previous studies have explored the rate of global warming, with some suggesting a potential slowdown or acceleration in recent decades. The influence of El Niño-Southern Oscillation (ENSO) events and other internal climate variations on observed temperature anomalies has been a subject of ongoing research. Methods used to address this internal variability include running means, multi-regression approaches using ENSO indices, and formal detection and attribution techniques that rely on forcing estimates and dedicated model simulations. The authors' approach offers a different perspective by using a model-based transfer function to physically link SST fluctuations to their global mean effect on GMST. This method is superior to simple running means because it considers underlying physical mechanisms, and it avoids the reliance on external forcing estimates inherent in other methods.

The authors utilize four major gridded surface air temperature datasets: HadCRUT5, GISTEMP v4, NOAA v5.1, and Berkeley Earth. They employ a model-based transfer function, derived from a 40-year CESM1.2.1-CAM5.3 Earth System Model simulation, to filter out the influence of SST patterns on GMST. This transfer function relates SST fluctuations in various ocean regions to their effect on global mean temperatures. The detrended SST patterns are used to calculate the total influence of ocean variability on the observed GMST anomalies, which are then removed from the raw data. This approach allows for a clearer separation of the anthropogenic warming signal from the effects of internal variability. The filtered data are then used to calculate 50-, 20-, and 10-year warming rates using linear regressions. These results are compared with those obtained from the unfiltered data to assess the effect of the filtering procedure. Additionally, the study compares the observed warming rates and rate changes to those simulated by an ensemble of 119 CMIP6 Earth System Models, using both historical emissions and Shared Socioeconomic Pathways (SSPs) to project future changes. The models' outputs are also filtered using the same transfer function as the observed data.

The analysis reveals a remarkably consistent 50-year (1973-2022) warming rate of approximately 0.18 °C/decade across all four datasets, even after filtering for SST influence. The inclusion of the recent La Niña-influenced years (2021 and 2022) did not alter this long-term trend significantly. However, a pronounced increase in the warming rate is observed since around 1990. When using a 20-year sliding window analysis, a warming rate increase of 0.016 °C/decade/decade is found across all four datasets. This increase is robust across window lengths exceeding 20 years. The raw data also indicate a warming rate increase but show greater uncertainty. CMIP6 models, on average, show higher 50-year warming rates and lower rate increases compared to the observations. Models with higher Equilibrium Climate Sensitivities (ECS) tend to exhibit a greater rate increase in warming, but even these high-ECS models don't fully capture the combination of observed long-term warming rate and recent acceleration. The observed data lies outside of the CMIP6 ensemble variability range, which presents challenges for climate risk assessments and impact studies that rely on CMIP6 model projections. The strongest increase in warming rates appears as a step-up around the mid-point of the 50-year period, rather than a continuous acceleration.

The finding of a steady long-term warming rate, coupled with a significant acceleration since 1990, highlights the complexities of climate change. The observed warming rate increase since 1990 is not fully explained by CMIP6 models. This discrepancy may be due to several factors, including deficiencies in the representation of Pacific SSTs in the models, incomplete understanding of aerosol forcing, and the climate response to volcanic eruptions like Mount Pinatubo's 1991 eruption. Further research is needed to refine our understanding of these factors and improve model representation. The upcoming El Niño event is expected to lead to record-high GMST values, emphasizing the continuing urgency of climate action. Future studies should incorporate a wider range of factors, multiple models and more sophisticated analyses, including consideration of stratospheric temperatures, volcanic eruptions, and the heterogeneous pattern of anthropogenic aerosol emissions, to enhance climate risk assessment and inform policymaking. The methodology employed, while robust, relies on a single Earth System Model to develop the Green's function. Future work could benefit from incorporating Green's functions from multiple models and exploring the sensitivity of the results to different model choices.

This study demonstrates a consistent 50-year warming rate, while also detecting a significant rate increase since 1990. This observed pattern is not well replicated by CMIP6 models, implying potential uncertainties in current climate projections. The study emphasizes the importance of continued monitoring of GMST and the need for improved model representation of key climate processes to accurately assess and mitigate climate risks.

The primary limitation of this study is the use of a single Earth System Model (CESM1.2.1-CAM5.3) to derive the Green's function used for filtering SST influence. While the robustness of the findings was tested, the use of multiple models could enhance the reliability of the results and improve the generalizability. Additionally, the study focuses primarily on SST fluctuations, excluding other potentially significant factors like stratospheric temperatures and volcanic eruptions. Finally, uncertainties in the aerosol radiative forcing estimates remain, affecting the accuracy of both the observational data and CMIP6 model simulations.