Assessing the impact of observations on a local numerical fog prediction system

As poor visibility conditions have great influence on air traffic, a need exists for accurate, updated fog and low-cloud forecasts. COBEL-ISBA, a boundary-layer one-dimensional numerical model, has been developed for the very short-term forecasting of fog and low clouds. This forecasting system assimilates the information from a local observation system designed to provide details on the state of the surface boundary layer, as well as that of the fog and low-cloud layers. This article aims to assess the influence of each component of the observation system on the initial conditions and low-visibility forecasts. The objective is to obtain a quantitative assessment of the impact on numerical fog forecasts of using a reduced (for smaller-sized airports) or enhanced (using a sodar) set of observations. We first used simulated observations, and focused on modelling the atmosphere before fog formation and then on simulating the life-cycle of fog and low clouds. Within this framework, we also estimated the impact of using a sodar to estimate the thickness of the cloud layer. We showed that the radiative flux observations were the most important of all in cloudy conditions, and that the measurement mast did not have to be higher than 10 m. Using either a sodar or radiative flux to estimate the optical thickness of a cloud layer gave the same scores. Using both of them together did not significantly improve the forecast. Simulations with real observations over a winter of simulations confirmed these findings. Copyright (C) 2009 Royal Meteorological Society