we are having supervised and unsupervised algorithms using neural networks to forecast using remote sensing and metrological data. Non of them have designed the algorithm
I know that you can use the meteorological models to forecast frost for the limited period. These models are usually used by the meteorological researchers.
Dear Elahe, I'm not sure what are you asking. Are you interested in frost identification and nowcasting or are you interested in incorporation of remote sensing data into a NWP with special interest in frost forecasting improvement?
For frost nowcasting, during the clear day/night it is possible to identify large scale frost from satellite micro-physical channels where frost and snow exhibit some similar spectral features. Recently we had that case. It is not excellent tool since it is impossible to distinguish land snow from frost, and it requires clear sky, but it is nice to have something like that. For more information you could search for Daniel Rosenfeld and his work on the RGB Microphysics algorithm for the EUMETSAT MSG geostationary satellites.
For frost forecasting, I do not see how remove sensing would help, I have some feeling that if you are dealing with NWP you will be more interested in ground measurements than remote sensing due to the errors of remote sensing data, but there are others here who might have different ideas.