Rodney J. Potts

Fog Forecasting for Melbourne Airport Using a Bayesian Decision Network

AbstractFog events occur at Melbourne Airport, Melbourne, Victoria, Australia, approximately 12 times each year. Unforecast events are costly to the aviation industry, cause disruption, and are a safety risk. Thus, there is a need to improve operational fog forecasting. However, fog events are difficult to forecast because of the complexity of the physical processes and the impact of local geography and weather elements. Bayesian networks (BNs) are a probabilistic reasoning tool widely used for prediction, diagnosis, and risk assessment in a range of application domains. Several BNs for probabilistic weather prediction have been previously reported, but to date none have included an explicit forecast decision component and none have been used for operational weather forecasting. A Bayesian decision network [Bayesian Objective Fog Forecast Information Network (BOFFIN)] has been developed for fog forecasting at Melbourne Airport based on 34 years’ worth of data (1972–2005). Parameters were calibrated to ens...

A simple inversion method for determining optimal dispersion model parameters from satellite detections of volcanic sulfur dioxide

A simple inversion scheme for optimizing volcanic emission dispersion model parameters with respect to satellite detections is presented in this paper. In this scheme, multiple dispersion model simulations, obtained by varying relevant model parameters, are created and compared against satellite detections using pattern correlation as a measure of model agreement with observations. It is shown that the scheme is successful in inferring emission source parameters such as those describing the vertical extent of the nascent sulfur dioxide emissions in the November 2010 Mount Merapi eruption in Java, Indonesia. These optimal parameter values then become a basis for improved forecasts of the transport of volcanic emissions.

Estimation of optimal dispersion model source parameters using satellite detections of volcanic ash

In this paper we demonstrate how parameters describing the geometry of the volcanic ash source for a particular volcanic ash dispersion model (HYSPLIT) may be inferred by the use of satellite data and multiple trial simulations. The areas of space likely to be contaminated by ash are identified with the aid of various remote sensing techniques and polygons are drawn around these areas as they would be in an operational setting. Dispersion model simulations are initialized either by a cylindrical source or a specified ash distribution depending on the context. Parameters of interest such as the base and top height, diameter, and optimal release time of the cylindrical source or the height of the specified ash distribution are inferred by forming a parameter grid and running multiple simulations for each parameter grid-point value. Optimal values of the parameter values are identified by calculating spatial correlations between the model simulations and observations. We demonstrate that the methodology can be used to correctly infer various model parameters and improves volcanic ash forecasts in various eruption case studies.

Toward quantitative forecasts of volcanic ash dispersal: Using satellite retrievals for optimal estimation of source terms

The provision of reliable quantitative forecasts for volcanic ash, such as ash mass load fields, is challenging because ash emission characteristics at the volcanic source are poorly understood. In this paper we show how satellite retrievals of volcanic ash mass load may be used to estimate source terms for dispersion models. The source terms comprise the spatial, temporal, and particle size distribution of mass flux at the ash source. We approach the problem by specifying general functional forms for these quantities that are dependent on a limited number of parameters. Numerous trial dispersion model simulations are then run, each corresponding to a particular configuration of possible source term parameters, with the resulting simulated mass load matched against the satellite retrieved mass load. The parameter values leading to best matches between simulations and satellite retrievals are then used to provide optimal forecasts of volcanic ash mass load distribution. We use several case studies to demonstrate the efficacy of this approach in improving forecasts of ash mass load with the HYSPLIT dispersion model.