3D Radiative Heating of Tropical Upper Tropospheric Cloud System Derived from Synergistic A-Train Observations and Machine Learning

Abstract:

Upper tropospheric (UT) clouds, most abundant in the tropics, play a crucial role in modulating the Earth’s energy budget and heat transport. They often originate from convective outflow, building large anvils. Yet the relation between these anvils and convection is not understood. The complete description of these systems and their atmospheric heating is a critical component to advance our knowledge, because the atmospheric heating gradients considerably influence the atmospheric circulation, which itself then modifies the patterns of intense precipitation.

The Infrared Sounders AIRS and IASI provide cloud properties with a large instantaneous horizontal coverage. Thus UT cloud systems can be identified by adjacent measurements of similar cloud height. Cloud emissivity is then used to distinguish convective cores, thick and thin cirrus anvils. This Cloud System Concept allows to link anvil properties to convection. Radiative heating rates are derived from CALIPSO lidar and CloudSat radar measurements of the cloud vertical structure, but only on narrow nadir tracks.

We built 3D radiative heating fields from 2003 to 2018 by applying machine learning techniques on cloud properties from IR sounders and atmospheric properties from meteorological reanalyses. The neural network regression models were trained on four years of collocated A-Train data. Together with the Cloud System approach we present the impact of tropical mesoscale convective systems. Tropicswide UT clouds enhance the column-integrated latent heating by about 25%. However, in the warmer regions the radiative heating profiles are nearly completely driven by deep convective cloud systems, which are colder and include slightly more thin cirrus around their anvils than those in cooler regions. They heat the whole troposphere from 200 hPa downward, while the LW cooling above the convective cores and thick cirrus anvils is opposed by thin cirrus heating. We also present the 15-year time series of the heating / cooling effects in relation to ENSO variations. The relative coverage of cold convective systems significantly increases with surface warming by 18 ± 5 %/K. The stronger heating of deeper convective systems seems to be counterbalanced by slightly stronger cooling above low-level clouds within the tropical and subtropical latitude band.