Although fog is an important weather phenomenon, it remains difficult to predict using traditional methods. This could be improved by new observations-based nowcasting systems. It has long been understood that fog affects measurements of atmospheric electricity. However, there has been disagreement in the literature on whether these changes contain information which is valuable for fog prediction beyond other commonly used methods. Here, results are presented which show that the potential gradient (PG), a measure of atmospheric electricity, could be used as an additional diagnostic in predicting fog for timescales of several hours. A much larger dataset of fog and PG is examined than has been previously possible, which allows for a more robust understanding of the behaviour of the PG during radiation fog. It is found to increase by a median of 58 V/m by the start of the event. In addition, this increase is found to begin over two hours in advance of the fog, 30% of the time. This shows that PG may contain useful fog nowcasting information. A number of individual fog case studies are presented and the applicability of the general results to these specific cases is discussed.