Application of advanced data assimilation techniques to the study of cloud and precipitation feedbacks in the tropical climate system

Abstract

The research documented in this study centers around two topics: examination of the response of precipitating cloud systems to changes in the tropical climate system, and assimilation of cloud and precipitation information from remote-sensing platforms on cloud resolving scales. The motivation for this work proceeds from the following outstanding problems: 1. Use of models to study the response of clouds to perturbations in the climate system is hampered by uncertainties in cloud microphysical parameterizations. 2. Though there is an ever-growing set of available observations, cloud and precipitation assimilation remains a difficult problem, particularly in the tropics. 3. Though it is widely acknowledged that cloud and precipitation processes play a key role in regulating the Earth's response to surface warming, the response of the tropical hydrologic cycle to climate perturbations remains largely unknown. The three above issues are addressed in a two part study. First, uncertainties in the NASA Goddard Cumulus Ensemble (GCE) model's cloud microphysical scheme are assessed and constrained using Bayesian estimation (data assimilation) methods. Because cloud microphysical parameters are highly nonlinearly related to the cloudy state, traditional data assimilation methods, which are based on a quasi-linear Gaussian-statistics estimation framework are not suitable. Instead, a Markov chain Monte Carlo (MCMC) algorithm, which makes no assumptions on the character of the state space or the linearity of the parameter-state relationship, is used to build a sample of the joint conditional parameter probability space, and to characterize the sensitivity of simulation results to changes in cloud microphysical parameters. It is found that simulations are most sensitive to the parameters that define the rain drop size distribution and the graupel and snow fall velocities, and it is concluded that TRMM retrieved mean precipitation rates, cloud properties, and radiative fluxes and heating rates can be effectively used to constrain these parameters to values consistent with convection in the tropics. In the second component of this study, the parameter values returned by MCMC are implemented in the GCE model, and the resulting observation-constrained model is used to investigate the response of cloud and precipitation processes to changes in sea surface temperature (SST). Two different experiments are performed--one in which base-state winds are set equal to zero, and one in which base-state winds are set equal to mean flow conditions typical of the active phase of the monsoon over the equatorial western Pacific. In each set of experiments, the SST is fixed at values of 295 K, 300 K, and 305 K, and the characteristics of the resulting equilibrium state are examined for evidence of consistent changes to the tropical hydrologic cycle with increasing SST. It is found that changes to SST alter the mode and scale of the convective organization for both sets of experiments, and that the manner in which convection organizes on large scales is fundamentally different between sheared and unsheared simulations. In addition, with increasing SST, the spatial and temporal mean cloud fraction decreases, the low cloud amount increases, the outgoing longwave radiation and column integrated latent heat release increase, and the shortwave flux at the top of the atmosphere decreases. At the same time, consistent with an increase in latent heating, the total surface precipitation increases, as does the precipitation intensity and efficiency. In spite of the complex interaction between the large scale circulation, convection, and cloud microphysical processes, increases in convective precipitation efficiency are found to be uniformly associated with decreases in high cloud fraction and decreases in outgoing shortwave radiation.