This paper addresses the problem of reliably predicting an important HF communication systems parameter, the critical frequency of the F2 ionospheric layer, with the use of a new machine learning technique, called Conformal Prediction (CP). CP accompanies the predictions of traditional machine learning algorithms with measures of confidence. The proposed approach is based on the wellknown Ridge Regression technique, but instead of the point predictions produced by the original method, it produces predictive intervals that satisfy a given confidence level. Our experimental results on an extended critical frequency dataset show that the obtained intervals are well-calibrated and narrow enough to be useful in practice.