Response of tropical convection to the presence of Saharan dust over the Atlantic Ocean during summer 2013

Kurzfassung

Tropical convection over the Atlantic Ocean is subject to entrainment of and interactions with desert dust from Africa during the summer season. Satellite data (i. a. of the Aerosol_cci project) are used to simultaneously assess dust transport over the tropical Atlantic Ocean as well as properties of convective clouds and cirrus in the basin during summer 2013. Infrared observations with the IASI instrument are independent of solar illumination and thus provide two observation cycles per day resulting in an increased sampling rate compared to solar instruments. These observations are used together with rainfall observations from the Tropical Rainfall Measuring Mission for a statistical analysis of the impact of soil dust on the probability distributions of cloud properties. The spatial distribution of dust AOD correlation length over the Atlantic Ocean basin is calculated for assessing the maximum distance at which dust observations can be assumed to impact on cloud properties. Principal component analysis of an observation vector including dust AOD, ice cloud properties and rainfall then shows several modes of co-variability of dust AOD and cloud development. The principal components reveal that the most dominant mode of cloud property variability (explaining 37% of the variance) is related to the cloud top temperature and is hardly connected to dust AOD. The second dominant mode (25% explained variance) shows significant co-variability between dust AOD and cloud /rainfall properties. It points to a strong increase of cloud optical depth and rainfall together with a smaller decrease of ice crystal size and cloud ice fraction at elevated dust levels. Higher order eigenvectors explain the co-variability of cloud top temperature and – more important – rainfall with dust AOD. The magnitudes of co-variability change slightly when only observations with precipitation are taken into account. Modes of co-variability as found through the principal component analysis are based on correlation and thus do not indicate an estimate for a cause-impact relationship. Therefore, and for quantification of the relationships, further statistical analysis is performed. A Bayesian method is used to estimate the contribution of varying meteorological conditions to the changes in distributions of cloud properties under varying dust levels. Consequently the impact of dust on the properties of ice clouds is assessed by this method. The abovementioned principal component analysis suggests that constraining the meteorological boundary conditions by observations of cloud top temperature in the Bayesian analysis scheme is well justified. The main findings of the Bayesian analysis are: a significant increase of ice cloud optical depth by 25% (moderate dust load) to 41% (high dust load) on average with increasing dust level. Joint histograms of cloud optical depth and dust AOD furthermore show that this increase is found to be a rather steady function of dust AOD. Moreover also the asymmetry of the probability density distribution of ice cloud optical depth increases under elevated dust levels as found by several parameters (skewness parameter, distribution-quartiles). Thus very high ice cloud optical depths are more likely to be observed near dust plumes than under pristine conditions. The Bayesian analysis scheme also shows that ice crystal sizes are decreases (by about 1-2% on average) under dusty conditions, exactly as the second principal component suggests. In addition to the decrease of the average crystal size, the probability distribution for the ice cloud effective radius also gets significantly narrower when the dust load increases. The most significant changes in ice cloud properties occur at cloud top temperatures between 260K and 240K, where it is assumed that the ability of dust particles serving as efficient ice nuclei induces early freezing of cloud droplets. This is also directly seen from an increase in cloud top ice fraction at these temperatures, especially for high dust loads. Rainfall is suppressed by about 15% on average under elevated dust levels. The optical depth increase and effective crystal size decrease are amplified if only rainy observations are analyzed. This result is also supported by the principal component analysis of the observation vectors of precipitating clouds. Although rainfall is suppressed in general the analysis shows that it is amplified for cloud systems with high optical depth and low cloud top temperature, i.e. deep convective systems. Consequently, rainfall becomes less frequent but stronger in precipitating convective systems as a result of dust interactions. The results of this statistical analysis imply a significant interaction between dust transport within the Saharan Air Layer and convective activity over the tropical Atlantic Ocean. They showcase the different ways how dust can affect convective clouds including intensification of convective activity. The asymmetric impact on the probability density distributions reveals an increase in severe convection, as observation of strong convection becomes more likely at elevated dust levels. Although it is typically assumed that dust reduces the likelihood of tropical storms due to increased wind shear at dust fronts, this analysis of co-variability between dust AOD and cloud parameters also shows the possibility of invigoration of severe convection. These effects may eventually lead to the formation or strengthening of tropical storms or counteract the wind shear effects.