The impact of parametrized convection on cloud feedback

Abstract

We investigate the sensitivity of cloud feedbacks to the use of convective parametrizations by repeating the CMIP5/CFMIP-2 AMIP/AMIP + 4K uniform sea surface temperature perturbation experiments with 10 climate models which have had their convective parametrizations turned off. Previous studies have suggested that differences between parametrized convection schemes are a leading source of inter-model spread in cloud feedbacks. We find however that 'ConvOff' models with convection switched off have a similar overall range of cloud feedbacks compared with the standard configurations. Furthermore, applying a simple bias correction method to allow for differences in present-day global cloud radiative effects substantially reduces the differences between the cloud feedbacks with and without parametrized convection in the individual models. We conclude that, while parametrized convection influences the strength of the cloud feedbacks substantially in some models, other processes must also contribute substantially to the overall inter-model spread. The positive shortwave cloud feedbacks seen in the models in subtropical regimes associated with shallow clouds are still present in the ConvOff experiments. Inter-model spread in shortwave cloud feedback increases slightly in regimes associated with trade cumulus in the ConvOff experiments but is quite similar in the most stable subtropical regimes associated with stratocumulus clouds. Inter-model spread in longwave cloud feedbacks in strongly precipitating regions of the tropics is substantially reduced in the ConvOff experiments however, indicating a considerable local contribution from differences in the details of convective parametrizations. In both standard and ConvOff experiments, models with less mid-level cloud and less moist static energy near the top of the boundary layer tend to have more positive tropical cloud feedbacks. The role of non-convective processes in contributing to inter-model spread in cloud feedback is discussed.

Keywords: climate; cloud; convection; feedback; parametrization.

Figure 1.

(a) Global mean net cloud feedbacks in CFMIP amip/amip4K experiments and convoffamip/convoffamip4K experiments without parametrized convection. This is diagnosed as the change in the global mean net cloud radiative effect (CRE) between the amip and amip4K experiments, normalized by the global mean near-surface temperature response and includes the effects of climatological cloud masking on the non-cloud feedbacks. Black lines denote the ranges in the values and the diagonal line indicates the one-to-one line. The lengths of the vertical coloured lines indicate the differences between standard and ConvOff values for the individual models. The linear correlation coefficient r is also shown. Panel (b) shows the same but with the ConvOff feedbacks rescaled by the factor required to bring the global mean net CRE in the convoffamip experiment into agreement with the standard amip experiment. Panel (c) shows the result of scaling all feedbacks by the factors required to bring their control experiments into agreement with an observed value of the net CRE (−17.1 W m⁻²).

Figure 2.

Scatterplot of LTS and precipitation from the HadGEM2-A amip experiment for February 1979 over the low-latitude oceans (30°N/S). The angular LTS/precipitation index (ALPI) is diagnosed as the angle of declination of a line connecting each point in LTS/precipitation space with an ‘anchor point’ on the top right. Locations with the strongest precipitation rates give values of ALPI of around 5°, whereas locations with the largest values of LTS result in an ALPI value of around 85°. Grey lines indicate the boundaries of ALPI percentile bins each covering 10% of the low-latitude ocean area.

Figure 3.

Composites of net, shortwave and longwave cloud radiative effect (CRE) over low-latitude oceans (30°N/S) in the amip control experiments (a,c,e) and convoffamip (b,d,f), sorted by percentiles of the angular LTS/precipitation index (ALPI). Black diamonds denote correlations with the net cloud feedback in the same ALPI bin which are significant at the 95% level. Squares indicate a significant correlation with the values in the bin and the average of the net cloud feedback over the entire low-latitude ocean domain. Ensemble mean values are shown with a black dashed line.

Figure 4.

As figure 3 but for maximum low, mid and high cloud fractions. These are dimensionless, taking values between 0 and 1, and are diagnosed from profiles of monthly mean cloud fraction on model levels by taking the maximum values in the pressure ranges 0–440, 440–680 and 680 hPa–surface.

Figure 5.

As figure 3 but for liquid water path (LWP) and ice water path (IWP). The compressed ranges of the LWP and IWP scales are chosen to support comparison of the smaller values, but by necessity exclude the 0–10th percentile values of LWP for the CNRM-CM5 convoffamip experiment (1.0 mm) and the 0–10th percentile values of IWP for GFDL AM2 and HIRAM convoffamip experiments (0.47 and 0.51 mm, respectively).

Figure 6.

Composites of net, shortwave and longwave cloud feedback over low-latitude oceans (30°N/S) in the amip/amip4K experiments (a,c,e) and convoffamip/convoffamip4K experiments (b,d,f), sorted by percentiles of the angular LTS/precipitation index (ALPI). Regions of strongest precipitation associated with deep convection fall in the lower percentiles while regions of strong static stability where shallow clouds predominate fall into the higher percentiles. The black dashed line shows the ensemble mean values in each bin.

Figure 7.

As figure 3 but for moist static energy (MSE) at 700 and 850 hPa.