| FluxNet is the neural network version of Streamer, a radiative transfer model. Given a set of input data consisting of surface, cloud, and atmospheric characteristics, FluxNet calculates upwelling and downwelling surface flux in either shortwave or longwave. While it is not as flexible as Streamer, FluxNet is faster by two to four orders of magnitude, making it ideal for large batch jobs and image processing. Four versions of the network are currently available: two shortwave and two longwave, one of each requiring temperature and humidity profiles in the input stream. Both were developed for global conditions.
FluxNet is available as executable programs for both Microsoft Windows and UNIX machines, and as an IDL subroutine. The source code, in C, is also available and should be platform-independent.
Caution: It is important to note that special versions of FluxNet may need to be designed and implemented for particular applications. For example, a special version was developed for use with the the ISCCP data because of differences in the input variable types and characteristics (e.g., ranges). Instructions for designing your own network are given in the User's Guide. |
