Does global warming amplify interannual climate variability?

Abstract

Based on the outputs of 30 models from Coupled Model Intercomparison Project Phase 5 (CMIP5), the fractional changes in the amplitude interannual variability (σ) for precipitation (P′) and vertical velocity (ω′) are assessed, and simple theoretical models are constructed to quantitatively understand the changes in σ(P′) and σ(ω′). Both RCP8.5 and RCP4.5 scenarios show similar results in term of the fractional change per degree of warming, with slightly lower inter-model uncertainty under RCP8.5. Based on the multi-model median, σ(P′) generally increases but σ(ω′) generally decreases under global warming but both are characterized by non-uniform spatial patterns. The σ(P′) decrease over subtropical subsidence regions but increase elsewhere, with a regional averaged value of 1.4% K^− 1 over 20°S–50°N under RCP8.5. Diagnoses show that the mechanisms for the change in σ(P′) are different for climatological ascending and descending regions. Over ascending regions, the increase of mean state specific humidity contributes to a general increase of σ(P′) but the change of σ(ω′) dominates its spatial pattern and inter-model uncertainty. But over descending regions, the change of σ(P′) and its inter-model uncertainty are constrained by the change of mean state precipitation. The σ(ω′) is projected to be weakened almost everywhere except over equatorial Pacific, with a regional averaged fractional change of − 3.4% K^− 1 at 500 hPa. The overall  reduction of σ(ω′) results from the increased mean state static stability, while the substantially increased σ(ω′) at the mid-upper troposphere over equatorial Pacific and the inter-model uncertainty of the changes in σ(ω′) are dominated by the change in the interannual variability of diabatic heating.

Appendix A

Suppose a variable X is proportional to the product of the variables Y and Z, i.e., X = kYZ where k is a non-zero constant. In a warmer climate, all of these three variable will change. If we denote the mean state in 20C with an overbar and the absolute change in 21C relative to 20C with a prefix of $\Delta$, the relationship between these three variables for 21C is expressed as

$$\bar {X}+\Delta X=k\left( {\bar {Y}+\Delta Y} \right)\left( {\bar {Z}+\Delta Z} \right)=k\bar {Y}\,\bar {Z}+k\bar {Y}\Delta Z+k\bar {Z}\Delta Y+k\Delta Y\Delta Z$$

(13)

Since $\bar {X}=k\bar {Y}\,\bar {Z}$ holds for the 20C, and the high-order term $\Delta Y\Delta Z$ is generally small and negligible, Eq. (13) can be simplified into

$$\Delta X=k\bar {Y}\Delta Z+k\bar {Z}\Delta Y$$

(14)

Since $\bar {X}=k\bar {Y}\,\bar {Z}$, if the left-hand side of Eq. (14) is divided by $\bar {X}$ and the right-hand side of Eq. (14) is divided by $k\bar {Y}\,\bar {Z}$, the following relation is obtained

$$\Delta X/\bar {X}~=\Delta Y/\bar {Y}~+~\Delta Z/\bar {Z}~~$$

(15)

If the prefix “$\delta$” to adopted to denote the relative change of a variable to its climatology in 20C, Eq. (15) can be expressed as

$$\delta X=\delta Y+\delta Z$$

(16)

In all, if variable X is proportional to the product of the variables Y and Z, the percentage change in X under global warming is the sum of the percentage changes in Y and Z. Similarly, if the variable Z is proportional to the quotient between X and Y, i.e., Z = kX/Y, the relationship $\delta Z=\delta X - \delta Y$ can also be obtained.