Quantifying the radiative response to surface temperature variability: A critical comparison of current methods

TL;DR

The paper tackles how the radiative response $R$ to surface-temperature forcing depends on the spatial pattern of warming (the pattern effect) and compares a suite of methods from Green's function experiments to statistically derived approaches. It systematically assesses each method's ability to predict the global radiative response from temperature patterns, using ~1000 years of piControl data for training and testing on internal variability as well as a $4\times CO_2$ forcing scenario. Internally, all methods capture substantial variance (roughly $50\%-80\%$) in $R$, but disagree on regional feedback signs and spatial structure, and differ markedly in forced-response magnitudes due to linearity assumptions and patch-size effects. The work highlights how different methods illuminate distinct facets of the pattern effect, discusses observational prospects, and outlines pathways to integrate these approaches for improved physical understanding of radiative feedbacks under pattern-dependent forcing.

Abstract

Over the past decade, it has become clear that the radiative response to surface temperature change depends on the spatially varying structure in the temperature field, a phenomenon known as the "pattern effect". The pattern effect is commonly estimated from dedicated climate model simulations forced with local surface temperatures patches (Green's function experiments). Green's function experiments capture causal influences from temperature perturbations, but are computationally expensive to run. Recently, however, several methods have been proposed that estimate the pattern effect through statistical means. These methods can accurately predict the radiative response to temperature variations in climate model simulations. The goal of this paper is to compare methods used to quantify the pattern effect. We apply each method to the same prediction task and discuss its advantages and disadvantages. Most methods indicate large negative feedbacks over the western Pacific. Over other regions, the methods frequently disagree on feedback sign and spatial homogeneity. While all methods yield similar predictions of the global radiative response to surface temperature variations driven by internal variability, they produce very different predictions from the patterns of surface temperature change in simulations forced with increasing CO2 concentrations. We discuss reasons for the discrepancies between methods and recommend paths towards using them in the future to enhance physical understanding of the pattern effect.

Figures (5)

• Figure 1: Comparison of methods, part 1.
• Figure 1: Comparison of methods, part 2 (caption on following page).
• Figure 1: (Previous page.) Comparison of methods: For each method ordered approximately from simplest to most complex process: 1) The estimated sensitivity from the global radiative response to temperature at each location, given in units of milliWatts per square meter per Kelvin (magnitudes are not comparable from one map to another due to differing resolution and masking); 2) the radiative response predicted by each method in a 4xCO$_2$ warming experiment plotted against the simulated truth; 3) a summary of the formulae or procedures used; 4) advantages of the method; and 5) disadvantages of the method.
• Figure 2: Performance of each method at predicting $\hat{R}(t)$ in 100 years of piControl internal variability withheld from the training data, shown as time series of predictions vs. simulated truth (left) and as regressions of predicted values onto simulated values (right). Squared correlation coefficients for the 100-year test are given for each regression. The black dashed lines indicate 1:1, which would imply a perfect prediction.
• Figure 3: Comparison of Green’s function experiments performed with the patch-size protocol used in Alessi2023-ag (top row) vs. with a smaller patch size in the equatorial Pacific (bottom row). a) Average gradient for the convolutional neural net trained on internal variability, identical to the CNN sensitivity map in Fig. \ref{['fig:table']}. b) Patches outlined at the half-patch width for the global protocol, c) sensitivity map produced by using the trained CNN to predict radiative responses to patch perturbations following the Green’s function protocol, performed on the patches in (b), and d) sensitivity map produced from patch perturbations in MPI-ESM1.2. The map in (d) is identical to the Green’s function sensitivity map in Fig. \ref{['fig:table']}. e) Smaller patches outlined at the half-patch width for a new experiment limited to the equatorial Pacific. f,g) As in (c,d), but using the patch perturbations outlined in (e). Note the difference in scale resulting from patch size.