Surrogate cloud fields

• My homepage
  
• Articles
  

Surrogate cloud fields share certain, typically statistical, properties with real cloud field. Surrogate cloud fields are necessary as we are not able to measure a full 3-dimensional cloud, but still would like to work with cloud fields that are as close as possible to real measured clouds.

These pages present two algorithms to generate surrogate cloud fields. The first is a simple and fast iterative method which produces clouds with the linear autocorrelations (power spectrum) and the PDF of the measurements. The page on this method includes many example cloud fields, the programs and articles in pdf-format.
The second is an evolutionary search algorithm that searches for a cloud which fits with the (measured) statistics in a cost function. As this cost function can contain (almost) arbitrary statistics, this is a very flexible method, but it is computationally much more expensive.

Furthermore, you can find here a text on the structure of cloud fields, and in how far this has a fractal nature. At the moment many researchers are working on generating synthetic cloud fields. I made a list of some of the classical and many current ones.

Further reading

• The simple and fast iterative method
• The flexible, but expensive constrained method
• A blog post on the fractal and non-fractal structure of cloud fields.
• An overview of current and much used cloud generators.

With the iterative method, you can make 3-dimensional surrogate clouds that have the cloud water distribution and the power spectrum of a measured cloud. For example, shown here, a 2D field from a 1D measurement.

With a global search algorithm, you can make constrained surrogate clouds that can have any (statistical) property that can be efficiently described by a cost function. A flexible, but computationally expensive method.