DETECTION AND MODELLING OF FOG USING ARTIFICIAL NEURAL NETWORK: A CASE STUDY OF LAHORE

Abstract

This study develops an artificial neural network (ANN) model to estimate fog occurrence in Lahore using daily records of day, month, minimum temperature, dew point, relative humidity, wind speed, and visibility for the November–March seasons from 2014 to 2017. The dataset was divided equally into training and testing subsets (50% each), and a randomly selected 40% of the data were also used for validation. The ANN architecture comprised seven input variables, a single hidden layer with 10 neurons, and one output representing fog occurrence. Model estimates were compared with observed fog-occurrence values using an absolute-difference threshold of 0.05. The comparisons indicate that the model reproduces most of the observed values reasonably well. These findings suggest that an ANN can be a useful approach for estimating fog occurrence in Lahore.