The estimation of a climatology of in-flight icing conditions has been a goal of the icing research community for a number of years. In-flight icing occurs when supercooled liquid water droplets impact on an aircraft during flight, forming a layer of ice on forward surfaces. The handling characteristics of the aircraft can be dramatically affected, resulting in a loss of lift in the most extreme cases. Much effort has been expended to develop forecasts of icing conditions aloft, but little is known about the distribution of icing over the continental United States. However, development of a climatology of icing conditions is not straightforward, due to the limited observations of this phenomenon that are available. In particular, widespread observations of icing are only available in the form of pilot reports (PIREPs) from aircraft. Unfortunately, these observations are sporadic and non-systematic, and cannot be used directly to provide a coherent or meaningful measure of the frequency of icing conditions in many locations. Thus, a statistical approach is taken in the development of a model for the icing climatology, using observations available in the regions of large cities. Icing reports are expected to be nearly systematic in those regions, because the air traffic is relatively frequent, and is consistent from day to day. Surface climatological variables have been found to be related to the icing observations in these locations, and these variables are used to develop a statistical model of icing frequency. The resulting model is applied to airport locations serving smaller populations, using their local surface climatological observations to provide estimates of the true frequencies of icing conditions. The resulting frequency estimates provide a more coherent and consistent climatology of icing than obtained from the PIREPs alone. Methods to quantify the uncertainty in the estimates are also investigated and applied.