A Multi-Variable Approach to Diagnosing the Monthly Covariability of the Amazonian Radiative and Convective Diurnal Cycles

The diurnal cycle of convection (CDC) greatly influences the water, radiative, and energy budgets in convectively active regions. For example, previous research of the Amazonian CDC has identified significant monthly covariability between the satellite-observed radiative and precipitation diurnal and multiple reanalysis-derived atmospheric state variables (ASVs) representing convective instability. However, disagreements between retrospective analysis products (reanalyses) over monthly ASV anomalies create significant uncertainty in the resulting covariability. Satellite observations of convective clouds can be used to characterize monthly anomalies in convective activity. CloudSat observes multiple properties of both deep convective cores and the associated anvils, and so is useful as an alternative to the use of reanalyses. CloudSat cannot observe the full diurnal cycle, but it can detect differences between daytime and nighttime convection. Initial efforts to use CloudSat data to characterize convective activity showed that the results are highly dependent on the choice of variable used to characterize the cloud. This is caused by a series of inverse relationships between convective frequency, cloud top height, radar reflectivity vertical profile, and other variables. A single, multi-variable index for convective activity based on CloudSat data may be useful to clarify the results. Principal component analysis (PCA) provides a method to create a multivariable index, where the first principal component (PC1) corresponds with convective instability. The time series of PC1 can then be used as a proxy for monthly variability in convective activity. The primary challenge presented involves determining the utility of PCA for creating a robust index for convective activity that accounts for the complex relationships of multiple convective cloud variables, and yields information about the interactions between convection, the convective environment, and radiation beyond the previous single-variable approaches. The choice of variables used to calculate PC1 may influence any results based on PC1, so it is necessary to test the sensitivity of the results to different variable combinations.