Data Requirements from Evolvable Sensor Networks for Homeland Security Problems Sue Ellen Haupt Applied Research Laboratory & Meteorology Department The Pennsylvania State University haupts2@asme.org Kerrie J. Long Applied Research Laboratory & Meteorology Department The Pennsylvania State University kjl203@psu.edu George S. Young Meteorology Department The Pennsylvania State University young@meteo.psu.edu Anke Beyer Meteorology Department The Pennsylvania State University aub166@psu.edu Abstract In the event of a release of toxic contaminant, either accidental or intentional, it would be useful to have an evolvable sensor network for tracking the toxic plume. In such an event, Homeland Security or DoD personnel are responsible for modeling the transport and dispersion of the plume. To do this requires specific source and meteorological data. Such data may not be available; however, it might be recoverable from concentration data monitored by a mobile sensor network. To be useful for assimilating the monitored data into a dispersion model, the sensor network must be sited strategically and should be evolvable to follow the plume of toxic contaminant. This paper discusses the requirements of such a network from the point of view of data needs for assimilating the sensor data into the transport and dispersion models. 1. Introduction It is incumbent upon Homeland Security and DoD officials to accurately predict the transport and dispersion of a toxic contaminant that may be released either accidentally or intentionally. To accurately predict the transport and dispersion requires detailed information about the source characteristics (time, location, height, amount, and type of release) plus meteorological data specific to the locale and time [1]. Such data are seldom available. An alternative approach is to place sensor networks in the vicinity of the release and use the monitored data to infer the appropriate source and meteorological data via data assimilation and inverse modeling. Such data assimilation approaches require a sufficient temporal and spatial density of sensor data to provide adequate concentration fields. Ideally, there would be a large number of sensors available to monitor the transport and dispersion of the contaminant. Such ideal networks are unlikely to be available. Instead, a fleet of unmanned aerial vehicles (UAVs) could be used to carry sensors to the most appropriate locations to provide sufficient information to resolve the plume. The purpose of this paper is to discuss the issues impacting data requirements for siting and controlling a fleet of UAVs for monitoring toxic concentrations in order to back-calculate the characteristics of the release and to assimilate the sensor data into the transport and dispersion model. The difficulty with accurately predicting transport and dispersion of a toxic contaminant stems from various causes of uncertainty. Specifically: • We often have inexact or incomplete source information. • The meteorological data may not be appropriate to the specific locale of the release. • Monitored concentration data contains errors. • There are inherent uncertainties in modeling turbulent dispersion. • Transport and dispersion models compute the ensemble average of many realizations of an event Second NASA/ESA Conference on Adaptive Hardware and Systems(AHS 2007) 0-7695-2866-X/07 $25.00 © 2007