This document describes the development of an enhanced support vector regression model for weather forecasting with more uninterpretable kernel functions. The model uses a two-stage process, first classifying weather parameters using self-organizing maps and then predicting weather events using support vector machines with multiple kernel functions. Experimental results show the enhanced model with kernel functions improves prediction accuracy compared to support vector machines alone or multilayer perceptrons. Future work could compare the model to other approaches or use it with weather image and hurricane data.