Many tropical island ecosystems are dependent on fog as a water input, but fog deposition dynamics across space in time in these systems are often poorly understood. In this study, we utilize ground-based cameras to develop the first ever non-binary fog classification system using in-situ cameras. We use this novel fog index to track fog cover at three different elevations in Ascension Island, UK. We then utilize a random forest model to predict fog cover using elevation, time and weather variables. Our data suggest that simple models such as random forest and logistic regression can be used to predict degree-of-fogginess with commonly measured meteorological variables. We also found that dewpoint depression was the single strongest predictor of fog cover, but the correlation between fog cover and other meteorological variables vary by elevation. This suggests that fog formation mechanisms likely vary by elevation, even along a relatively short transect. In assessing these variables with the diurnal patterns of fog cover at different elevations, we suggest that in the lower valley, radiation fog was the dominant fog type and in the higher elevations, advection/orographic fog were likely the dominant fog types. Overall, this work demonstrates that our novel fog tracking method can potentially be utilized to track fog in other remote, data-scarce tropical regions.