Maximum rates of climate change are systematically underestimated in the geological record

Abstract

Recently observed rates of environmental change are typically much higher than those inferred for the geological past. At the same time, the magnitudes of ancient changes were often substantially greater than those established in recent history. The most pertinent disparity, however, between recent and geological rates is the timespan over which the rates are measured, which typically differ by several orders of magnitude. Here we show that rates of marked temperature changes inferred from proxy data in Earth history scale with measurement timespan as an approximate power law across nearly six orders of magnitude (10(2) to >10(7) years). This scaling reveals how climate signals measured in the geological record alias transient variability, even during the most pronounced climatic perturbations of the Phanerozoic. Our findings indicate that the true attainable pace of climate change on timescales of greatest societal relevance is underestimated in geological archives.

Figure 1. Magnitudes and rates of geological and recent temperature changes plotted against measurement timespan.

(a) Magnitudes of published oceanic, continental (cont.) and recent temperature changes. Black line is the linear regression slope through the combined ocean and continental geological data (0.10). Dashed lines define the 95% uncertainty envelope of the scaling relationship based on our Monte Carlo error analysis (that is, 0.09±0.03, see main text and Methods). Data associated with well-known warming events are highlighted in orange: B/A=Bølling–Allerød warming (∼13.7 kyr), H=Last Glacial Maximum-Holocene transition (∼11.7 kyr), P/E=Paleocene–Eocene thermal maximum (∼56 Myr ago), 1b=early Albian OAE1b (∼110 Myr ago), eT=early Toarcian (∼182 Myr ago), T/J=Triassic-Jurassic boundary (∼201 Myr ago), P/Tr=Permian–Triassic boundary (∼252 Myr ago). The two points for the P/Tr (a,b) are from refs. , , respectively, discussed in Introduction. (b) Data plotted as rates of temperature change against timespan, highlighting negative power law relationship in continental and oceanic data. The scaling relationship of the combined ocean and continental geological data is –0.90 (black line). As in a, dashed black lines define the 95% uncertainty envelope derived from our Monte Carlo error analysis (−0.91±0.03). See Supplementary Data 1 for full data listing.

Figure 2. Magnitudes of sea surface and polar temperature changes.

Timespan-dependence of sea surface temperature (SST) change is plotted along with maximum magnitudes (heavy black line, see Methods), not including the extreme values of the Permian–Triassic boundary event (orange points). The coloured cluster density plot shows the distribution of >1.6 million temperature changes calculated from an ensemble of 27 single-site SST records, binned into 0.5 × 0.05 log bins (see Methods and Supplementary Table 1). Also shown are maximum magnitudes of temperature changes from the ensemble (grey line) and the Antarctica Dome-C (ref. 24) and Greenland GISP2 (ref. 25) temperature records (red dashed lines).

Figure 3. Timespan-corrected climate change rates.

Rates were standardized by removing the pervasive timespan-dependent scaling trend in the compilation of continental (cont.) and oceanic data (slope=−0.90; Fig. 1) and normalizing to the mean rate at timespans between 1,000 and 2,000 years (dashed line). This timespan range represents the maximum temporal resolution likely achievable in most palaeoclimate records (see Methods and ref. 12). The normalization emphasizes how the fastest rates of geological climate change were likely attained across the Permian–Triassic boundary event (P/Tr^a and P/Tr^b, ∼252 Myr ago, see also Supplementary Fig. 1b).