Stronger zonal convective clustering associated with a wider tropical rain belt

Deep convection can exhibit a large diversity of spatial organizations along the equator. The form of organization may affect the tropical large-scale motions of the atmosphere, but observational evidence is currently missing. Here we show using observations that when convection along the equator is more clustered in the zonal direction, the tropical rain belt widens in the meridional direction, and exhibits a double-peak structure. About half of the influence of the convective clustering on the width of the rain belt is associated with the annual cycle and the other half is associated with unforced climate variability. Idealized climate model experiments show that the zonal convective clustering alone can explain the observed behavior and that the behavior can be explained with an energetic framework. This demonstrates that the representation of equatorial convective clustering is important for modeling the tropical rainfall distribution accurately.

Supplementary Figure 1 | Spatial correlation between evaporation and precipitation. The figure shows the spatial and daily correlation coefficient squared between the meridional mean evaporation and the precipitation between 6 S and 6 N as a function of the amplitude of the imposed evaporation forcing (at the equator) normalized by the equatorial mean (S λ (E)). The two simulations with amplitude 0 are the control simulation and the simulation with no zonal variations.
Supplementary Figure 2 | Radiative cooling in observations. Panel a) shows the total vertically integrated atmospheric radiative cooling, panel b) the atmospheric cloud-radiative effect (ACRE) averaged from 6 S to 6 N as a function of the zonal convective clustering (S λ (P )) for the GPCP observational dataset from March 2000 to December 2016. The radiative cooling tendencies were calculated from the CERES-EBAF data set. Red circles denote the months with a tropical precipitation distribution that is symmetric about the equator (see methods for details) and black circles those with an asymmetric distribution. The red lines are obtained by linear regression to the red circles and the black lines are obtained by linear regression to all (red and black) circles.
Supplementary Table 1 | Relationship between the width and the double-peak structure of the intertropical convergence zone. The table shows the variance explained between the precipitation-inferred width of the intertropical convergence zone (ITCZ, W P ), the dynamically inferred width of the ITCZ (W ω ) and the meridional distance between the two zonal-mean peaks in precipitation (φ S ). all indicates that the variance explained was calculated using all months, sym using only the months with a precipitation distribution that is symmetric about the equator (see methods) and Simulations using the statistically steady-state mean of the simulations. Note that if there is only one peak in precipitation, the distance between the two peaks is set to zero. The sign indicates if the correlation is positive or negative.
W P W ω Amplitude (+) 50% (+) 62% Zonal wavenumber (-) 23% (-) 17% Supplementary Table 2 | Relation between the imposed evaporation forcing width of the intertropical convergence zone The table shows the variance explained between the amplitude as well as the zonal wavenumber of the imposed evaporation forcing in the aquaplanet simulations and the precipitation-inferred (W P ) as well as the dynamically-inferred (W ω ) width of the intertropical convergence zone. The control simulation was excluded from the calculation of the variance explained, because no forcing was imposed. The sign indicates if the correlation is positive or negative.
Supplementary Table 4 | Relationship between different metrics of zonal convective clustering. The table shows the variance explained between the zonal standard deviation of precipitation normalized by the mean precipitation (S λ (P )), the mean area fraction of subsidence (F ω500>0 ) and the minimum fraction of surface (and time) necessary to accumulate 80% of the total precipitation (F 0.8P ), all calculated in the region from 6 S to 6 N. The sign indicates if the correlation is positive or negative. all indicates that the variance explained was calculated using all months, sym using only the months with a precipitation distribution that is symmetric about the equator (see methods) and Simulations using the statistically steady-state mean of the simulations.