Assessment of Forecast Skill of High- and Coarse-Resolution Numerical Weather Prediction Models in Predicting Visibility/ Fog Over Delhi, India

Accurate forecasts of visibility are important to avoid disruption in air and highway traffic caused due to the formation of dense fog. However, accurate forecasting of visibility/fog remains a challenge as the genesis and development of fog is a result of many processes. In view of this, models have been developed in recent years to forecast visibility and the occurrence of fog is measured in terms of visibility. The global Unified Model of the National Centre for Medium Range Weather Forecasting, known as NCUM, provides direct output of visibility. As aviation is severely affected at the Indira Gandhi International Airport, New Delhi, India, a high-resolution model was set up to forecast visibility over the airport. The present study analyses the performance of the coarse-resolution global model and high-resolution model in predicting visibility over Delhi. Visibility is categorized into three ranges – very poor (0–200 m), poor (200–1000 m) and clear conditions beyond 1 km. The accuracy of forecast in different ranges of visibility is determined using different statistical scores. Evaluation of the results shows that the performance of both high and coarse resolution model remains low in poor visibility conditions. Though the high-resolution model performs better than the coarse-resolution model in predicting a drop in visibility, it also has higher number of false alarms. None of the model is able to predict the very poor visibility conditions. The prediction of visibility from the high-resolution model can further be improved by inclusion of real-time aerosol fields in the model.