STARneuro combines the radiative transfer model with the technique of neural networks. The radiative transfer is solved only for a few wavelengths and by neural network technique the full high-resolution spectrum is replenished, leading to a strong reduction of computation time compared to STARsci. The limitations concerning the accuracy of the calculated spectrum is less than 2.5% for wavelengths higher than 310 nm and less than 10% below. Detailed information can soon be found in:

Schwander, H., A. Kaifel, A. Ruggaber, and P. Koepke, Spectral radiative transfer modelling with minimized computation time using neural network technique., Appl. Opt., Vol. 40, No. 3, p. 331-335, 2001.

The technique of neural networks is also used to calculate UV global irradiance for all kinds of cloudiness. Depending on the information, available to describe the cloud conditions, STARneuro is able to calculate global irradiance spectra with an accuracy of up to 15% for conditions with broken cloudiness. Compared to STARsci the detector geometries that can be chosen are reduced:

•    global irradiance, no direct and diffuse component  

•    integrated radiation quantities based on global irradiance (with or without weighting by action spectra)  

•    time integrations (dose rates)  

The calculations are carried out for variable heights in the troposphere above ground. An algorithm to deal with specific ground height above sea level is included. The atmospheric conditions for the model runs are specified simply by mouse-click within a JAVA interface. Own data sets for wavelength field, spectral albedo, vertical profiles of atmospheric constituents, etc. can aesily be provided by the user.