A Python-based Parcel Model Framework for Studying Aerosol-Cloud Processes

2016-07-08T19:26:20Z (GMT) by Daniel Rothenberg
<h2>Adiabatic parcel theory provides a simple framework for studying fundamental aerosol-cloud-precipitation interactions. Numerical models based on this theory are commonly used within the field to (a) perform droplet closure studies with in situ aerosol and cloud observations from aircraft, (b) develop parameterizations of aerosol-cloud interactions for global models, and (c) study the sensitivity of cloud processes to changes in the ambient aerosol distribution. Additionally, they can accomodate a range of microphysical schemes ranging from simple to highly detailed and can be modified to simulate many conditions from fixed heating to variable entrainment rates.<p>However, because parcel models are of an intermediate software complexity (they are more sophisticated than crude thermodynamics calculators, but far less complicated than cloud resolving or single-column models) they often serve as throw-away code which is not rigorously developed, tested, and documented. This limits the reproducibility and verifiability of results obtained with them, because the scientific and numerical formulation of a particular parcel model scheme can significantly impact the characteristics of its simulations. To help alleviate this, a parcel model code used to study droplet activation has been modified to serve as an adaptable parcel modeling framework that can be easily modified or extended to reflect different scientific use cases. The model is implemented in an object-oriented manner so that many instances of it can easily be run simultaneously on large clusters from a single, simple script. Furthermore, it utilizes Cython to optimize performance-critical numerical calculations, provides full documentation using Sphinx and Python docstrings, and implements a novel, Monte Carlo particle-based approach for simulating warm cloud microphysics. The code is wholly open-source and hosted on GitHub with a permissive BSD license; conda packages are provided for building on Linux and Mac OS X.</p><p>To highlight its flexibility, results from using the model framework to both build a new droplet activation scheme for global models and to analyze field campaign observations of urban aerosol CCN properties are presented. Prospects for future development - particularly an extension to study IN impacts on cirrus or mixed-phase microphysics - are also discussed </p></h2>