Large scale simulations of cloud droplet growth in turbulent environments
Title: Large scale simulations of cloud droplet growth in turbulent environments
SNIC Project: SNIC 2014/1-47
Project Type: SNAC Medium
Principal Investigator: Rodrigo Caballero <>
Affiliation: Stockholms universitet
Duration: 2014-03-01 – 2015-03-01
Classification: 10508 20306


Clouds play a fundamental role in the dynamics of the atmosphere and the water on the earth. Our knowledge of cloud dynamics is still so poor that it represents the cause of a relevant uncertainty in climate predictions and atmospheric circulation models. Our project investigates the dynamics of liquid water drops in warm clouds interacting with the surrounding turbulent flow. These processes are difficult to capture because of the multi-scale nature of the turbulent flows at the high Reynolds numbers typical of a cloud. Although the characteristic length scale of the droplets (microns) is smaller than the smallest active turbulence scale (Kolmogorov scale: millimeters), it is known that the droplet size distribution is strongly affected by turbulence. In particular, turbulence is responsible of the broadening of the droplet-size spectra that promotes collisions between droplets of different sizes and consequently rain formation. Nowadays it is impossible to capture the complete dynamics of a cloud both in laboratory experiments and direct numerical simulation (DNS). In addition, it is also impossible to measure directly these properties in a real cloud because of the limited resolution and control of the experimental devices. The aim of the project is to quantify the water droplet dynamics in a small portion of a real cloud (order meters) by means of direct numerical simulation where the complete and full droplet-turbulence interactions will be analyzed in details. Direct Numerical Simulations (DNS) is a powerful numerical technique able to solve turbulent flows in all details, from the largest to the smallest scales without any approximation. DNS have now been employed for almost 30 years and data are usually in perfect agreement with theoretical predictions and experiments. This technique is an "ideal experiment" as it gives access to instantaneous and statistical quantities not possible to measure even with the most modern and accurate instrumental systems. Nevertheless, the DNS does have limitations: need of huge numbers of grid points and of big supercomputers, long simulation times and large data storages, possibility to simulate cases with moderate-small turbulence levels (Reynolds number). Owing to these limitations, our simulations will be combined with upscale models to be able to extend and interpret the results at the physical scale of a real cloud, (characterized by higher Reynolds number). A novel multi-scale methodology has been specifically designed to give an accurate prediction of the droplet-size spectra broadering induced by turbulence. The DNS simulations will be performed by means of a pseudo-spectral Navier-Stokes solver coupled with an equation for the supersaturation field and a Lagrangian solver for the droplet dynamics. The code is fully parallelized to be efficiently used on supercomputing infrastructures. The simulations we shall perform will produce state of the art data of turbulent-droplets and multiphase flow dynamics. The computational grid will involve 1 billion grid points for the Eulerian fields (3 velocity and the supersaturation) and 1 billion of discrete droplet trajectories will be integrated.