The intentional or accidental release of airborne toxics poses great risk to the public health. During these incidents, the greatest factor of uncertainty is related to the location and rate of released substance, therefore, an information of high importance for emergency preparedness and response plans. A novel computational algorithm is proposed to estimate, efficiently, the location and release rate of an airborne toxic substance source based on health effects observations; data that can be readily available, in a real accident, contrary to actual measurements. The algorithm is demonstrated by deploying a semi-empirical dispersion model and Monte Carlo sampling on a simplified scenario. Input data are collected at varying receptor points for toxics concentrations (C; standard approach) and two new types: toxic load (TL) and health effects (HE; four levels). Estimated source characteristics are compared with scenario values. The use of TL required the least number of receptor points to estimate the release rate, and demonstrated the highest probability (>90%). HE required more receptor points, than C, but with lesser deviations while probability was comparable, if not better. Finally, the algorithm assessed very accurately the source location when using C and TL with comparable confidence, but HE demonstrated significantly lower confidence.
Read full abstract