The Feasibility of Using Large Ensembles to Identify Human Influence on the Seasonal Cycle of Atmospheric Temperature

Benjamin Santer | 21-FS-035

Project Overview

In a paper published in Science in 2018, Lawrence Livermore National Laboratory scientists analyzed T{AC}, the geographical pattern of the size of the annual cycle of tropospheric temperature. A human fingerprint in T{AC} was detectable in satellite temperature data. This fingerprint had distinctive features: a larger seasonal cycle over mid-latitudes, a reduced seasonal cycle over high latitudes, and small changes in the tropics. Model simulations showed that greenhouse gas increases from fossil fuel burning were the primary cause of these changes. This LDRD feasibility study built on this prior research, addressing two major questions. First, was the result reported on in the Science paper a robust result—a result that was relatively insensitive to uncertainties in natural climate variability? Second, can idealized simulations help scientists to better understand the physical mechanisms responsible for the pattern of changes in the seasonal temperature cycle? We used five different large initial condition ensembles (LEs) to address question 1. In an LE, every ensemble member is generated with the same climate model and the same external influences (such as historical changes in greenhouse gases), but starts from different initial conditions of the climate system. Each ensemble member provides a unique realization of natural internal variability superimposed on the climate response to the external influences. One can use LEs to ask whether a human fingerprint on the seasonal temperature cycle is robustly identifiable in each realization. If it is, this enhances confidence that identification of a human-caused seasonal cycle fingerprint in satellite temperature data is due to anthropogenic factors, and not to the effects of multi-decadal internal variability (MIV). Livermore pioneered the use of LEs in climate fingerprint research, but had not used them for studying T{AC} changes.

To answer questions about the physical mechanisms responsible for seasonal cycle changes, idealized "aquaplanet" simulations (Earth-like simulations with continents removed) helped us to probe the impact of excluding land. These simulations had seasonally varying insolation under preindustrial and quadrupled CO2 conditions, and explored a range of different values of sea-ice reflectivity. The five LEs analyzed here relied on models with different applied historical external influences, sensitivity to greenhouse-gas (GHG) increases, and MIV properties. Despite these differences, their fingerprints of annual cycle changes were similar and could be identified with high statistical confidence in 239 of the 240 LE realizations of historical climate change. Comparing simulated and observed variability spectra revealed that consistent fingerprint identification was unlikely to be biased by significant model underestimates of observed MIV. Our results suggest that the positive identification of an anthropogenic annual cycle fingerprint in observations reflects a robust response to basic physical processes that are well-represented in climate models. MIV does not drive observed T{AC} changes.

The aquaplanet simulations captured the prominent mid-latitude increases in T{AC} evident in observations and in "human influence" simulations performed with full Earth System Models (ESMs). The aquaplanet runs did not capture the observed and ESM-simulated hemispheric asymmetry in these features (larger mid-latitude T{AC} increases in the Northern Hemisphere than in the Southern Hemisphere). The comparisons between aquaplanet runs, observations, and ESMs suggest that mid-latitude annual cycle increases are likely driven by seasonally dependent, GHG-induced changes in atmospheric circulation and tropospheric stability, while the hemispheric asymmetry in these increases is due to land–ocean differences in heat capacity and to summertime surface drying.

Mission Impact

The changes in seasonality identified here are likely to have significant socioeconomic impacts. Examples include impacts on fire weather, runoff, water supply, and agriculture. Ecosystem impacts may also be profound—the seasonal cycle is the pacemaker of many key processes in the biological world. It is therefore critically important to understand the physical mechanisms driving historical changes in the seasonal cycle of atmospheric temperature, and to make better projections of seasonal cycle behavior over the 21st century. The LDRD feasibility study has helped to maintain the leadership position of Livermore and DOE in climate change detection and attribution research, strengthened the Laboratory's existing scientific collaborations with remote sensing experts, climate modelers, and model diagnosticians, and established new collaborations with groups generating LEs and performing aquaplanet simulations. These collaborations will benefit Livermore and DOE efforts to produce and analyze LEs performed with the DOE's Energy Exascale Earth System Model (E3SM).

Publications, Presentations, and Patents

Santer, B.D., et al. 2020. "Changes in the seasonal cycle of atmospheric temperature in the CESM1 and CESM2 large ensembles." Virtual presentation, CESM2 Large Ensembles workshop, November 2020.