We have built an automated Satellite Images Forecasting System with Neuro-Fuzzy techniques. Firstly, Subtractive Clustering is applied on to a satellite image to extract the locations of the clouds. This is followed by Fuzzy C-Means Clustering which operates on the next satellite image, seeded with the cloud clusters of the previous image. With the matching of cloud clusters across successive images, cloud cluster velocities are deduced. Using a Generalized Regression Neural Network, we interpolate the cloud cluster velocities over the whole area of interest. Finally, the linear forecasting scheme then moves each cloud pixel in that satellite image according to the velocities of the past hour.