Roland Draxler

A decision support system to aid in the risk evaluation of airborne animal diseases was developed for Ireland. The system's primary objective is to assist in risk evaluation of the airborne spread of Foot and Mouth Disease (FMD). The operational system was developed by Met Éireann – the Irish Meteorological Service and CVERA (Centre for Veterinary Epidemiology and Risk Analysis), in co-operation with NOAA-ARL (National Oceanic and Atmospheric Administration – Air Resources Laboratory) and ECMWF (European Centre for Medium-Range Weather Forecasts). The infrastructure largely relies on the HYSPLIT dispersion model driven by both ECMWF meteorological forecasts for longer range simulations, and HARMONIE-AROME meteorological forecasts, a high resolution local area meteorological model, ideal for shorter range national emissions. Following on from previous work by the Bureau of Meteorology, Australia as well as the Australian Department of Agriculture, Fisheries and Forestry, further modifications were made to the HYSPLIT source code to improve the model's characterisation of the Foot and Mouth Disease virus. FMD is a highly infectious disease among cloven hoofed animals that can transmit via airborne means. Biological characteristics related to temperature, humidity, lifespan as well as atmospheric washout were all incorporated either through new or existing functionality of the dispersion model. Combining the model dispersion capabilities of HYSPLIT with a virus emission model and GIS mapping software with farmland zoning, the disease dispersion system becomes a powerful analysis and decision support tool. This airborne animal disease atmospheric dispersion system helps improve emergency preparedness, as well as aid confinement and eradication strategies for relevant Irish authorities, during a disease outbreak.

Currently, the National Oceanic and Atmospheric Administration (NOAA) National Weather Service (NWS) runs the HYSPLIT dispersion model with a unit mass release rate to predict the transport and dispersion of volcanic ash. The model predictions provide information for the Volcanic Ash Advisory Centers (VAAC) to issue advisories to meteorological watch offices, area control centers, flight information centers, and others. This research aims to provide quantitative forecasts of ash distributions generated by objectively and optimally estimating the volcanic ash source strengths, vertical distribution, and temporal variations using an observation-modeling inversion technique. In this top-down approach, a cost functional is defined to quantify the differences between the model predictions and the satellite measurements of column-integrated ash concentrations weighted by the model and observation uncertainties. Minimizing this cost functional by adjusting the sources provides the volcanic ash emission estimates. As an example, MODIS (Moderate Resolution Imaging Spectroradiometer) satellite retrievals of the 2008 Kasatochi volcanic ash clouds are used to test the HYSPLIT volcanic ash inverse system. Because the satellite retrievals include the ash cloud top height but not the bottom height, there are different model diagnostic choices for comparing the model results with the observed mass loadings. Three options are presented and tested. Although the emission estimates vary significantly with different options, the subsequent model predictions with the different release estimates all show decent skill when evaluated against the unassimilated satellite observations at later times. Among the three options, integrating over three model layers yields slightly better results than integrating from the surface up to the observed volcanic ash cloud top or using a single model layer. Inverse tests also show that including the ash-free region to constrain the model is not beneficial for the current case. In addition, extra constraints on the source terms can be given by explicitly enforcing no-ash for the atmosphere columns above or below the observed ash cloud top height. However, in this case such extra constraints are not helpful for the inverse modeling. It is also found that simultaneously assimilating observations at different times produces better hindcasts than only assimilating the most recent observations.

Currently NOAA's National Weather Service (NWS) runs the HYSPLIT dispersion model with a unit mass release rate to predict the transport and dispersion of volcanic ash. The model predictions provide information for the Volcanic Ash Advisory Centers (VAAC) to issue advisories to meteorological watch offices, area control centers, flight information centers, and others. This research aims provide quantitative forecasts of ash distributions generated by objectively and optimally estimating the volcanic ash source strengths, vertical distribution and temporal variations using an observation-modeling inversion technique. In this top-down approach, a cost functional is defined to mainly quantify the differences between model predictions and the satellite measurements of column integrated ash concentrations, weighted by the model and observation uncertainties. Minimizing this cost functional by adjusting the sources provides the volcanic ash emission estimates. As an example, MODIS (MOderate Resolution Imaging Spectroradiometer) satellite retrievals of the 2008 Kasatochi volcanic ash clouds are used to test the HYSPLIT volcanic ash inverse system. Because the satellite retrievals include the ash cloud top height but not the bottom height, three options for matching the model concentrations to the observed mass loadings are tested. Although the emission estimates vary significantly with different options the subsequent model predictions with the different release estimates all show decent skill when evaluated against the unassimilated satellite observations at later times. Among the three options, integrating over three model layers yields slightly better results than integrating from the surface up to the volcanic ash cloud top or using a single model layer. Inverse tests also show that including the ash-free region to constrain the model is not beneficial for the current case. In addition, extra constraints to the source terms can be given by explicitly enforcing ``no-ash'' for the atmosphere columns above or below the observed ash cloud top height. However, in this case such extra constraints are not helpful for the inverse modeling. It is also found that simultaneously assimilating observations at different times produces better hindcasts than only assimilating the most recent observations.

Both the root mean square error (RMSE) and the mean absolute error (MAE) are regularly employed in model evaluation studies. Willmott and Matsuura (2005) have suggested that the RMSE is not a good indicator of average model performance and might be a misleading indicator of average error, and thus the MAE would be a better metric for that purpose. While some concerns over using RMSE raised by Willmott and Matsuura (2005) and Willmott et al. (2009) are valid, the proposed avoidance of RMSE in favor of MAE is not the solution. Citing the aforementioned papers, many researchers chose MAE over RMSE to present their model evaluation statistics when presenting or adding the RMSE measures could be more beneficial. In this technical note, we demonstrate that the RMSE is not ambiguous in its meaning, contrary to what was claimed by Willmott et al. (2009). The RMSE is more appropriate to represent model performance than the MAE when the error distribution is expected to be Gaussian. In addition, we show that the RMSE satisfies the triangle inequality requirement for a distance metric, whereas Willmott et al. (2009) indicated that the sums-of-squares-based statistics do not satisfy this rule. In the end, we discussed some circumstances where using the RMSE will be more beneficial. However, we do not contend that the RMSE is superior over the MAE. Instead, a combination of metrics, including but certainly not limited to RMSEs and MAEs, are often required to assess model performance.

Both the root mean square error (RMSE) and the mean absolute error (MAE) are regularly employed in model evaluation studies. Willmott and Matsuura (2005) have suggested that the RMSE is not a good indicator of average model performance and might be a misleading indicator of average error and thus the MAE would be a better metric for that purpose. Their paper has been widely cited and may have influenced many researchers in choosing MAE when presenting their model evaluation statistics. However, we contend that the proposed avoidance of RMSE and the use of MAE is not the solution to the problem. In this technical note, we demonstrate that the RMSE is not ambiguous in its meaning, contrary to what was claimed by Willmott et al. (2009). The RMSE is more appropriate to represent model performance than the MAE when the error distribution is expected to be Gaussian. In addition, we show that the RMSE satisfies the triangle inequality requirement for a distance metric.