Paper Summary: Impact of Atmospheric Cloud Radiative Effects on Annular Mode Persistence in Idealized Simulations
As discussed in the previous post, cloud-circulation interactions are the source of much uncertainty in climate models. Part of the difficulty of studying these interactions is the lack of simplified/idealized modeling set-ups for building understanding. We need testbeds that simulate large-scale (1000s of km) circulations while still resolving (or parameterizing) small-scale cloud processes on scales of several km or less. This likely means compromising on the representation of the large-scale or on the representation of clouds.
In a new paper, David Vishny developed an idealized modeling framework for studying the effects of clouds on the internal variability of the mid-latitude circulation. Specifically, we were interested in how Atmospheric Cloud Radiative Effects (ACRE) impact the annular mode, the leading mode of variability of the mid-latitude atmosphere. The annular mode is a north-south oscillation of the jet’s position which occurs on time-scales of 10-20 days and is one of the main sources of predictability of the mid-latitude atmosphere. Mid-latitude clouds shift with this oscillation, and we wanted to investigate how the associated changes in atmospheric heating affect the structure and persistence of annular mode anomalies. This built on a previous study led by Casey Wall in which we re-examined the structure of the ACRE associated with annular modes in satellite data.
To investigate the impacts of ACRE on the annular mode, David took a widely used “dry” atmospheric modeling set-up (Held-Suarez) and added a zonal-mean heating field which depends on the strength of the zonal index, a measure of the strength and sign of annular mode anomalies (see panel a in Figure 1). In our sign convention, a positive zonal index means the jet is north of its usual position. David also added a constant (a) to the parameterization of the ACRE, which allowed him to investigate the linearity of the ACRE impacts. E.g., setting a = -1 flips the sign of the heating applied to the model. This set-up allowed us to investigate how the time-evolving ACRE forcing affects annular modes, but it could be used more generally to investigate how time-evolving heatings affect the mid-latitude atmosphere.
Figure 1: Zonal-mean ACRE heating added to the model. (b) Difference in zonal-mean temperature anomalies associated with the zonal index between the a = 1 and control simulations. (c) Autocorrelations of the zonal index in the five simulations (solid lines); the dashed line marks e−1 (d) Changes in e-folding timescale of the autocorrelation (i.e., persistence) for the ACRE-forced simulations (black bars), and contributions from eddies, friction and MMC (colored bars).
David compared 5 simulations: a control (a = 0) simulation, an a = 1 simulation (Earth-like), an a = -1 simulation and simulations with a = +/-3 (strong forcing). For the a > 0 cases, the persistence of annular mode anomalies decreased (panel c in Figure 1) – the flow became less predictable. For the a = -1 case the persistence didn’t change from the control, while for a = -3 the persistence increased somewhat. So ACRE seems to reduce the persistence of the zonal index in the most realistic case, though the effects are highly nonlinear.
To explain these responses, David separated out the persistence of the zonal index into two regimes: a “fast” regime, in which the zonal index autocorrelation declines rapidly because of atmospheric noise, and a “slow” regime, in which the autocorrelation declines more slowly due to a positive eddy feedback (discussed e.g., here). Here, we found that the changes in persistence due to the ACRE were all due to the slow regime, as the “noise” was more-or-less unaffected by the presence of the heating.
Digging into it further, we found that for a > 0 the ACRE weakens the low-level temperature gradients which generate baroclinic eddies, thus weakening the eddy feedback. This effect is partially mitigated by reduced friction, because the modes become more top-heavy, but the main effect is that weaker eddies lead to a weaker feedback and less persistent anomalies (Figure 1d). For a < 0 the results were more complicated because the changes in the eddy forcing didn’t project cleanly onto the eddy feedback in the control case. So even though the dynamics were disrupted, they only weakly affected the persistence. We don’t understand these nonlinearities well, but suspect they’re related to nonlinearities in the temperature response. The friction reduces persistence for a < 0 because the modes become more bottom-heavy.
So for the Earth-like case the story seems to be fairly straightforward, as ACRE reduces the persistence of annular mode anomalies. But still we don’t understand the nonlinearities in the response well, largely because we don’t understand how low-level heating generates low level temperature anomalies, nor how these temperature anomalies affect eddy generation. It would be interesting to study this further. Our results also emphasize the “indirect” role of ACRE on eddy generation, rather than its “direct” role in forcing changes in jet position, which was generally small in these runs. There’s an interesting contrast with recent work by Lu et al who found that clouds increase jet persistence in a comprehensive GCM. They mostly attributed this to changes in condensation, an effect not included in our model, and their set-up wasn’t able to isolate the indirect effect of ACRE, which we found to be key. Reconciling our work and theirs, and “bridging the hierarchy” would be a good next step.
