My institute is currently using BEAM as part of the processing of time-series of AATSR and MERIS satellite images. We are using these images to estimate various parameters for snow and ice in several projects. We observe that BEAM is a major bottleneck in our processing chains. BEAM is used to convert AATSR and MERIS images from the N1 format in acquisition geometry to the dimap format in, e.g., geographical coordinates.

As a comparison, we have written our own conversion routine in IDL (ittvis.com) for converting MODIS images from hdf files in acquisition geometry to ENVI/IDL files. This conversion software is at least a hundred times faster than BEAM. This is probably due to the fact that IDL has been desinged for fast access to image arrays through the use of index arrays, a feature not found in regular programming languages like C and Fortran.

While BEAM's processing time may not matter too much if one is only interested in converting a single MERIS scene, this is a major issue when processing, e.g., all available MERIS scenes for one winter season within the area of interest (in one particular project this is most of Finland).

Could you please give us some advice on how to reduce the processing time with BEAM?

Best regards,
Øivind

Please also read

Creating a GPF Graph

and

Bulk Processing with GPT

It often helps to increase the size of the tile cache in order to reach better performance. And when you do a map projection, it may be a good idea to use the SubsetOp to extract the region you are interested in before making the map projection

And you may want to try different tile sizes by changing



in the beam.config file. The above setting makes the tile width equal to the full width of the input images. You may also read this thread.